[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16774397#comment-16774397 ] Sean Mackrory commented on HADOOP-15999: FWIW I'm not seeing that test failure, but can't dig into it much myself right now. > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: Sub-task > Components: fs/s3 >Affects Versions: 3.1.0 >Reporter: Sean Mackrory >Assignee: Gabor Bota >Priority: Major > Attachments: HADOOP-15999.001.patch, HADOOP-15999.002.patch, > HADOOP-15999.003.patch, HADOOP-15999.004.patch, HADOOP-15999.005.patch, > HADOOP-15999.006.patch, out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[
https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16774210#comment-16774210
]
Hadoop QA commented on HADOOP-15999:
| (/) *{color:green}+1 overall{color}* |
\\
\\
|| Vote || Subsystem || Runtime || Comment ||
| {color:blue}0{color} | {color:blue} reexec {color} | {color:blue} 0m
19s{color} | {color:blue} Docker mode activated. {color} |
|| || || || {color:brown} Prechecks {color} ||
| {color:green}+1{color} | {color:green} @author {color} | {color:green} 0m
0s{color} | {color:green} The patch does not contain any @author tags. {color} |
| {color:green}+1{color} | {color:green} test4tests {color} | {color:green} 0m
0s{color} | {color:green} The patch appears to include 1 new or modified test
files. {color} |
|| || || || {color:brown} trunk Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 22m
46s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
39s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
26s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 0m
41s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
13m 41s{color} | {color:green} branch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 0m
55s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
25s{color} | {color:green} trunk passed {color} |
|| || || || {color:brown} Patch Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 0m
39s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
36s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javac {color} | {color:green} 0m
36s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
21s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 0m
37s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} whitespace {color} | {color:green} 0m
0s{color} | {color:green} The patch has no whitespace issues. {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
13m 34s{color} | {color:green} patch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 1m
2s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
24s{color} | {color:green} the patch passed {color} |
|| || || || {color:brown} Other Tests {color} ||
| {color:green}+1{color} | {color:green} unit {color} | {color:green} 4m
46s{color} | {color:green} hadoop-aws in the patch passed. {color} |
| {color:green}+1{color} | {color:green} asflicense {color} | {color:green} 0m
26s{color} | {color:green} The patch does not generate ASF License warnings.
{color} |
| {color:black}{color} | {color:black} {color} | {color:black} 62m 27s{color} |
{color:black} {color} |
\\
\\
|| Subsystem || Report/Notes ||
| Docker | Client=17.05.0-ce Server=17.05.0-ce Image:yetus/hadoop:8f97d6f |
| JIRA Issue | HADOOP-15999 |
| JIRA Patch URL |
https://issues.apache.org/jira/secure/attachment/12959626/HADOOP-15999.006.patch
|
| Optional Tests | dupname asflicense compile javac javadoc mvninstall
mvnsite unit shadedclient findbugs checkstyle |
| uname | Linux 13ce888fc3e5 4.4.0-139-generic #165~14.04.1-Ubuntu SMP Wed Oct
31 10:55:11 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux |
| Build tool | maven |
| Personality | /testptch/patchprocess/precommit/personality/provided.sh |
| git revision | trunk / a868f59 |
| maven | version: Apache Maven 3.3.9 |
| Default Java | 1.8.0_191 |
| findbugs | v3.1.0-RC1 |
| Test Results |
https://builds.apache.org/job/PreCommit-HADOOP-Build/15959/testReport/ |
| Max. process+thread count | 347 (vs. ulimit of 1) |
| modules | C: hadoop-tools/hadoop-aws U: hadoop-tools/hadoop-aws |
| Console output |
https://builds.apache.org/job/PreCommit-HADOOP-Build/15959/console |
| Powered by | Apache Yetus 0.8.0 http://yetus.apache.org |
This message was automatically generated.
> [s3a] Better support for out-of-band operations
> ---
>
> Key: HADOOP-15999
> URL:
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16774129#comment-16774129 ] Gabor Bota commented on HADOOP-15999: - uploaded v006 with the toStrings > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: Sub-task > Components: fs/s3 >Affects Versions: 3.1.0 >Reporter: Sean Mackrory >Assignee: Gabor Bota >Priority: Major > Attachments: HADOOP-15999.001.patch, HADOOP-15999.002.patch, > HADOOP-15999.003.patch, HADOOP-15999.004.patch, HADOOP-15999.005.patch, > HADOOP-15999.006.patch, out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16774057#comment-16774057 ] Steve Loughran commented on HADOOP-15999: - -1 I want the toString values of the exceptions to include the full file status of what is going wrong, as I showed in my example. The latest patch doesn't do it, and as the test is still failing for me, not enough for me even to begin to debug it. > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: Sub-task > Components: fs/s3 >Affects Versions: 3.1.0 >Reporter: Sean Mackrory >Assignee: Gabor Bota >Priority: Major > Attachments: HADOOP-15999.001.patch, HADOOP-15999.002.patch, > HADOOP-15999.003.patch, HADOOP-15999.004.patch, HADOOP-15999.005.patch, > out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[
https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16773995#comment-16773995
]
Hadoop QA commented on HADOOP-15999:
| (/) *{color:green}+1 overall{color}* |
\\
\\
|| Vote || Subsystem || Runtime || Comment ||
| {color:blue}0{color} | {color:blue} reexec {color} | {color:blue} 0m
24s{color} | {color:blue} Docker mode activated. {color} |
|| || || || {color:brown} Prechecks {color} ||
| {color:green}+1{color} | {color:green} @author {color} | {color:green} 0m
0s{color} | {color:green} The patch does not contain any @author tags. {color} |
| {color:green}+1{color} | {color:green} test4tests {color} | {color:green} 0m
0s{color} | {color:green} The patch appears to include 1 new or modified test
files. {color} |
|| || || || {color:brown} trunk Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 18m
54s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
33s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
22s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 0m
38s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
12m 21s{color} | {color:green} branch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 0m
42s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
24s{color} | {color:green} trunk passed {color} |
|| || || || {color:brown} Patch Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 0m
28s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
27s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javac {color} | {color:green} 0m
27s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
18s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 0m
33s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} whitespace {color} | {color:green} 0m
0s{color} | {color:green} The patch has no whitespace issues. {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
12m 50s{color} | {color:green} patch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 0m
48s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
22s{color} | {color:green} the patch passed {color} |
|| || || || {color:brown} Other Tests {color} ||
| {color:green}+1{color} | {color:green} unit {color} | {color:green} 4m
37s{color} | {color:green} hadoop-aws in the patch passed. {color} |
| {color:green}+1{color} | {color:green} asflicense {color} | {color:green} 0m
32s{color} | {color:green} The patch does not generate ASF License warnings.
{color} |
| {color:black}{color} | {color:black} {color} | {color:black} 55m 25s{color} |
{color:black} {color} |
\\
\\
|| Subsystem || Report/Notes ||
| Docker | Client=17.05.0-ce Server=17.05.0-ce Image:yetus/hadoop:8f97d6f |
| JIRA Issue | HADOOP-15999 |
| JIRA Patch URL |
https://issues.apache.org/jira/secure/attachment/12959589/HADOOP-15999.005.patch
|
| Optional Tests | dupname asflicense compile javac javadoc mvninstall
mvnsite unit shadedclient findbugs checkstyle |
| uname | Linux fe5bcf84428c 3.13.0-153-generic #203-Ubuntu SMP Thu Jun 14
08:52:28 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux |
| Build tool | maven |
| Personality | /testptch/patchprocess/precommit/personality/provided.sh |
| git revision | trunk / eedcc8e |
| maven | version: Apache Maven 3.3.9 |
| Default Java | 1.8.0_191 |
| findbugs | v3.1.0-RC1 |
| Test Results |
https://builds.apache.org/job/PreCommit-HADOOP-Build/15958/testReport/ |
| Max. process+thread count | 319 (vs. ulimit of 1) |
| modules | C: hadoop-tools/hadoop-aws U: hadoop-tools/hadoop-aws |
| Console output |
https://builds.apache.org/job/PreCommit-HADOOP-Build/15958/console |
| Powered by | Apache Yetus 0.8.0 http://yetus.apache.org |
This message was automatically generated.
> [s3a] Better support for out-of-band operations
> ---
>
> Key: HADOOP-15999
> URL:
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16773939#comment-16773939 ] Gabor Bota commented on HADOOP-15999: - In my latest patch I've added texts for all asserts. I'm not able to reproduce the issue with the test. All tests look happy to me. Shared my configs and logs offline to figure out what could be the issue. > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: Sub-task > Components: fs/s3 >Affects Versions: 3.1.0 >Reporter: Sean Mackrory >Assignee: Gabor Bota >Priority: Major > Attachments: HADOOP-15999.001.patch, HADOOP-15999.002.patch, > HADOOP-15999.003.patch, HADOOP-15999.004.patch, HADOOP-15999.005.patch, > out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[
https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16773301#comment-16773301
]
Steve Loughran commented on HADOOP-15999:
-
BTW I like assert statements to always include enough information in messages
to be able to make sense of the failure without having to look into the source
code to find the assert and then guess what could be up.
As an example:
{code}
Assert.assertNotEquals(rawFileStatus.getModificationTime(),
guardedFileStatus.getModificationTime());
{code}
to
{code}
Assert.assertNotEquals("Modification time of raw matches that of guarded
\nraw=" + rawFileStatus + " guarded=" + guardedFileStatus
rawFileStatus.getModificationTime(),
guardedFileStatus.getModificationTime());
{code}
Just imagine that you've seen a jenkins build fail, and all you have is that
assertion text. What information does it need to have to help you understand
what has gone wrong?
> [s3a] Better support for out-of-band operations
> ---
>
> Key: HADOOP-15999
> URL: https://issues.apache.org/jira/browse/HADOOP-15999
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3
>Affects Versions: 3.1.0
>Reporter: Sean Mackrory
>Assignee: Gabor Bota
>Priority: Major
> Attachments: HADOOP-15999.001.patch, HADOOP-15999.002.patch,
> HADOOP-15999.003.patch, HADOOP-15999.004.patch, out-of-band-operations.patch
>
>
> S3Guard was initially done on the premise that a new MetadataStore would be
> the source of truth, and that it wouldn't provide guarantees if updates were
> done without using S3Guard.
> I've been seeing increased demand for better support for scenarios where
> operations are done on the data that can't reasonably be done with S3Guard
> involved. For example:
> * A file is deleted using S3Guard, and replaced by some other tool. S3Guard
> can't tell the difference between the new file and delete / list
> inconsistency and continues to treat the file as deleted.
> * An S3Guard-ed file is overwritten by a longer file by some other tool. When
> reading the file, only the length of the original file is read.
> We could possibly have smarter behavior here by querying both S3 and the
> MetadataStore (even in cases where we may currently only query the
> MetadataStore in getFileStatus) and use whichever one has the higher modified
> time.
> This kills the performance boost we currently get in some workloads with the
> short-circuited getFileStatus, but we could keep it with authoritative mode
> which should give a larger performance boost. At least we'd get more
> correctness without authoritative mode and a clear declaration of when we can
> make the assumptions required to short-circuit the process. If we can't
> consider S3Guard the source of truth, we need to defer to S3 more.
> We'd need to be extra sure of any locality / time zone issues if we start
> relying on mod_time more directly, but currently we're tracking the
> modification time as returned by S3 anyway.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[
https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16773288#comment-16773288
]
Steve Loughran commented on HADOOP-15999:
-
-1
from {{mvn verify -Dtest=moo -Ds3guard -Ddynamodb -Dscale
-Dit.test=ITestS3GuardOutOfBandOperations }}
{code}
[INFO] ---
[INFO] T E S T S
[INFO] ---
[INFO] Running org.apache.hadoop.fs.s3a.ITestS3GuardOutOfBandOperations
[ERROR] Tests run: 12, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 29.064
s <<< FAILURE! - in org.apache.hadoop.fs.s3a.ITestS3GuardOutOfBandOperations
[ERROR]
testListingLongerLengthOverwriteAuthoritative(org.apache.hadoop.fs.s3a.ITestS3GuardOutOfBandOperations)
Time elapsed: 1.562 s <<< FAILURE!
java.lang.AssertionError: Values should be different. Actual: 1550689095000
at
org.apache.hadoop.fs.s3a.ITestS3GuardOutOfBandOperations.overwriteFileInListing(ITestS3GuardOutOfBandOperations.java:310)
at
org.apache.hadoop.fs.s3a.ITestS3GuardOutOfBandOperations.testListingLongerLengthOverwriteAuthoritative(ITestS3GuardOutOfBandOperations.java:183)
[INFO]
[INFO] Results:
[INFO]
[ERROR] Failures:
[ERROR]
ITestS3GuardOutOfBandOperations.testListingLongerLengthOverwriteAuthoritative:183->overwriteFileInListing:310->Assert.assertNotEquals:211->Assert.assertNotEquals:199->Assert.failEquals:185->Assert.fail:88
Values should be different. Actual: 1550689095000
[INFO]
{code}
> [s3a] Better support for out-of-band operations
> ---
>
> Key: HADOOP-15999
> URL: https://issues.apache.org/jira/browse/HADOOP-15999
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3
>Affects Versions: 3.1.0
>Reporter: Sean Mackrory
>Assignee: Gabor Bota
>Priority: Major
> Attachments: HADOOP-15999.001.patch, HADOOP-15999.002.patch,
> HADOOP-15999.003.patch, HADOOP-15999.004.patch, out-of-band-operations.patch
>
>
> S3Guard was initially done on the premise that a new MetadataStore would be
> the source of truth, and that it wouldn't provide guarantees if updates were
> done without using S3Guard.
> I've been seeing increased demand for better support for scenarios where
> operations are done on the data that can't reasonably be done with S3Guard
> involved. For example:
> * A file is deleted using S3Guard, and replaced by some other tool. S3Guard
> can't tell the difference between the new file and delete / list
> inconsistency and continues to treat the file as deleted.
> * An S3Guard-ed file is overwritten by a longer file by some other tool. When
> reading the file, only the length of the original file is read.
> We could possibly have smarter behavior here by querying both S3 and the
> MetadataStore (even in cases where we may currently only query the
> MetadataStore in getFileStatus) and use whichever one has the higher modified
> time.
> This kills the performance boost we currently get in some workloads with the
> short-circuited getFileStatus, but we could keep it with authoritative mode
> which should give a larger performance boost. At least we'd get more
> correctness without authoritative mode and a clear declaration of when we can
> make the assumptions required to short-circuit the process. If we can't
> consider S3Guard the source of truth, we need to defer to S3 more.
> We'd need to be extra sure of any locality / time zone issues if we start
> relying on mod_time more directly, but currently we're tracking the
> modification time as returned by S3 anyway.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[
https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16773248#comment-16773248
]
Hadoop QA commented on HADOOP-15999:
| (/) *{color:green}+1 overall{color}* |
\\
\\
|| Vote || Subsystem || Runtime || Comment ||
| {color:blue}0{color} | {color:blue} reexec {color} | {color:blue} 1m
0s{color} | {color:blue} Docker mode activated. {color} |
|| || || || {color:brown} Prechecks {color} ||
| {color:green}+1{color} | {color:green} @author {color} | {color:green} 0m
0s{color} | {color:green} The patch does not contain any @author tags. {color} |
| {color:green}+1{color} | {color:green} test4tests {color} | {color:green} 0m
0s{color} | {color:green} The patch appears to include 1 new or modified test
files. {color} |
|| || || || {color:brown} trunk Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 18m
36s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
36s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
22s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 0m
42s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
12m 23s{color} | {color:green} branch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 0m
42s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
24s{color} | {color:green} trunk passed {color} |
|| || || || {color:brown} Patch Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 0m
34s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
28s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javac {color} | {color:green} 0m
28s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
18s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 0m
36s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} whitespace {color} | {color:green} 0m
0s{color} | {color:green} The patch has no whitespace issues. {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
13m 13s{color} | {color:green} patch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 0m
58s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
20s{color} | {color:green} the patch passed {color} |
|| || || || {color:brown} Other Tests {color} ||
| {color:green}+1{color} | {color:green} unit {color} | {color:green} 4m
45s{color} | {color:green} hadoop-aws in the patch passed. {color} |
| {color:green}+1{color} | {color:green} asflicense {color} | {color:green} 0m
23s{color} | {color:green} The patch does not generate ASF License warnings.
{color} |
| {color:black}{color} | {color:black} {color} | {color:black} 56m 27s{color} |
{color:black} {color} |
\\
\\
|| Subsystem || Report/Notes ||
| Docker | Client=17.05.0-ce Server=17.05.0-ce Image:yetus/hadoop:8f97d6f |
| JIRA Issue | HADOOP-15999 |
| JIRA Patch URL |
https://issues.apache.org/jira/secure/attachment/12959468/HADOOP-15999.004.patch
|
| Optional Tests | dupname asflicense compile javac javadoc mvninstall
mvnsite unit shadedclient findbugs checkstyle |
| uname | Linux 15b3c65cf169 4.4.0-138-generic #164~14.04.1-Ubuntu SMP Fri Oct
5 08:56:16 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux |
| Build tool | maven |
| Personality | /testptch/patchprocess/precommit/personality/provided.sh |
| git revision | trunk / aa3ad36 |
| maven | version: Apache Maven 3.3.9 |
| Default Java | 1.8.0_191 |
| findbugs | v3.1.0-RC1 |
| Test Results |
https://builds.apache.org/job/PreCommit-HADOOP-Build/15950/testReport/ |
| Max. process+thread count | 306 (vs. ulimit of 1) |
| modules | C: hadoop-tools/hadoop-aws U: hadoop-tools/hadoop-aws |
| Console output |
https://builds.apache.org/job/PreCommit-HADOOP-Build/15950/console |
| Powered by | Apache Yetus 0.8.0 http://yetus.apache.org |
This message was automatically generated.
> [s3a] Better support for out-of-band operations
> ---
>
> Key: HADOOP-15999
> URL:
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[
https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16773201#comment-16773201
]
Gabor Bota commented on HADOOP-15999:
-
Thanks [~ste...@apache.org]. This was an issue when using LocalMS. I've fixed
the issue in the test with adding a line to {{setAllowAuthoritativeInFs}}:
{{realMs = guardedFs.getMetadataStore();}}
Tests are now running clean against ireland with dynamo, local, and none MSs
> [s3a] Better support for out-of-band operations
> ---
>
> Key: HADOOP-15999
> URL: https://issues.apache.org/jira/browse/HADOOP-15999
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3
>Affects Versions: 3.1.0
>Reporter: Sean Mackrory
>Assignee: Gabor Bota
>Priority: Major
> Attachments: HADOOP-15999.001.patch, HADOOP-15999.002.patch,
> HADOOP-15999.003.patch, HADOOP-15999.004.patch, out-of-band-operations.patch
>
>
> S3Guard was initially done on the premise that a new MetadataStore would be
> the source of truth, and that it wouldn't provide guarantees if updates were
> done without using S3Guard.
> I've been seeing increased demand for better support for scenarios where
> operations are done on the data that can't reasonably be done with S3Guard
> involved. For example:
> * A file is deleted using S3Guard, and replaced by some other tool. S3Guard
> can't tell the difference between the new file and delete / list
> inconsistency and continues to treat the file as deleted.
> * An S3Guard-ed file is overwritten by a longer file by some other tool. When
> reading the file, only the length of the original file is read.
> We could possibly have smarter behavior here by querying both S3 and the
> MetadataStore (even in cases where we may currently only query the
> MetadataStore in getFileStatus) and use whichever one has the higher modified
> time.
> This kills the performance boost we currently get in some workloads with the
> short-circuited getFileStatus, but we could keep it with authoritative mode
> which should give a larger performance boost. At least we'd get more
> correctness without authoritative mode and a clear declaration of when we can
> make the assumptions required to short-circuit the process. If we can't
> consider S3Guard the source of truth, we need to defer to S3 more.
> We'd need to be extra sure of any locality / time zone issues if we start
> relying on mod_time more directly, but currently we're tracking the
> modification time as returned by S3 anyway.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[
https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16773104#comment-16773104
]
Steve Loughran commented on HADOOP-15999:
-
-1
ran the tests myself & got some NPEs
{code}
[INFO] Tests run: 45, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 88.128
s - in org.apache.hadoop.fs.s3a.select.ITestS3Select
[INFO]
[INFO] Results:
[INFO]
[ERROR] Errors:
[ERROR]
ITestS3GuardOutOfBandOperations.testListingDeleteAuthoritative:196->deleteFileInListing:342
NullPointer
[ERROR]
ITestS3GuardOutOfBandOperations.testListingDeleteNotAuthoritative:203->deleteFileInListing:344
NullPointer
[ERROR]
ITestS3GuardOutOfBandOperations.testListingLongerLengthOverwriteAuthoritative:183->overwriteFileInListing:287
NullPointer
[ERROR]
ITestS3GuardOutOfBandOperations.testListingLongerLengthOverwriteNotAuthoritative:190->overwriteFileInListing:289
NullPointer
[ERROR]
ITestS3GuardOutOfBandOperations.testListingSameLengthOverwriteAuthoritative:170->overwriteFileInListing:287
NullPointer
[ERROR]
ITestS3GuardOutOfBandOperations.testListingSameLengthOverwriteNotAuthoritative:177->overwriteFileInListing:289
NullPointer
[INFO]
[ERROR] Tests run: 725, Failures: 0, Errors: 6, Skipped: 7
{code}
full detail
{code}
[ERROR]
testListingLongerLengthOverwriteNotAuthoritative(org.apache.hadoop.fs.s3a.ITestS3GuardOutOfBandOperations)
Time elapsed: 3.34 s <<< ERROR!
java.lang.NullPointerException
at
org.apache.hadoop.fs.s3a.ITestS3GuardOutOfBandOperations.overwriteFileInListing(ITestS3GuardOutOfBandOperations.java:289)
at
org.apache.hadoop.fs.s3a.ITestS3GuardOutOfBandOperations.testListingLongerLengthOverwriteNotAuthoritative(ITestS3GuardOutOfBandOperations.java:190)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at
sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
at
sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:498)
at
org.junit.runners.model.FrameworkMethod$1.runReflectiveCall(FrameworkMethod.java:50)
at
org.junit.internal.runners.model.ReflectiveCallable.run(ReflectiveCallable.java:12)
at
org.junit.runners.model.FrameworkMethod.invokeExplosively(FrameworkMethod.java:47)
at
org.junit.internal.runners.statements.InvokeMethod.evaluate(InvokeMethod.java:17)
at
org.junit.internal.runners.statements.RunBefores.evaluate(RunBefores.java:26)
at
org.junit.internal.runners.statements.RunAfters.evaluate(RunAfters.java:27)
at org.junit.rules.TestWatcher$1.evaluate(TestWatcher.java:55)
at
org.junit.internal.runners.statements.FailOnTimeout$CallableStatement.call(FailOnTimeout.java:298)
at
org.junit.internal.runners.statements.FailOnTimeout$CallableStatement.call(FailOnTimeout.java:292)
at java.util.concurrent.FutureTask.run(FutureTask.java:266)
at java.lang.Thread.run(Thread.java:745)
[ERROR]
testListingSameLengthOverwriteNotAuthoritative(org.apache.hadoop.fs.s3a.ITestS3GuardOutOfBandOperations)
Time elapsed: 1.694 s <<< ERROR!
java.lang.NullPointerException
at
org.apache.hadoop.fs.s3a.ITestS3GuardOutOfBandOperations.overwriteFileInListing(ITestS3GuardOutOfBandOperations.java:289)
at
org.apache.hadoop.fs.s3a.ITestS3GuardOutOfBandOperations.testListingSameLengthOverwriteNotAuthoritative(ITestS3GuardOutOfBandOperations.java:177)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at
sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
at
sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:498)
at
org.junit.runners.model.FrameworkMethod$1.runReflectiveCall(FrameworkMethod.java:50)
at
org.junit.internal.runners.model.ReflectiveCallable.run(ReflectiveCallable.java:12)
at
org.junit.runners.model.FrameworkMethod.invokeExplosively(FrameworkMethod.java:47)
at
org.junit.internal.runners.statements.InvokeMethod.evaluate(InvokeMethod.java:17)
at
org.junit.internal.runners.statements.RunBefores.evaluate(RunBefores.java:26)
at
org.junit.internal.runners.statements.RunAfters.evaluate(RunAfters.java:27)
at org.junit.rules.TestWatcher$1.evaluate(TestWatcher.java:55)
at
org.junit.internal.runners.statements.FailOnTimeout$CallableStatement.call(FailOnTimeout.java:298)
at
org.junit.internal.runners.statements.FailOnTimeout$CallableStatement.call(FailOnTimeout.java:292)
at java.util.concurrent.FutureTask.run(FutureTask.java:266)
at java.lang.Thread.run(Thread.java:745)
[ERROR]
testListingDeleteNotAuthoritative(org.apache.hadoop.fs.s3a.ITestS3GuardOutOfBandOperations)
Time elapsed: 1.44 s
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[
https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16773023#comment-16773023
]
Hadoop QA commented on HADOOP-15999:
| (/) *{color:green}+1 overall{color}* |
\\
\\
|| Vote || Subsystem || Runtime || Comment ||
| {color:blue}0{color} | {color:blue} reexec {color} | {color:blue} 0m
16s{color} | {color:blue} Docker mode activated. {color} |
|| || || || {color:brown} Prechecks {color} ||
| {color:green}+1{color} | {color:green} @author {color} | {color:green} 0m
0s{color} | {color:green} The patch does not contain any @author tags. {color} |
| {color:green}+1{color} | {color:green} test4tests {color} | {color:green} 0m
0s{color} | {color:green} The patch appears to include 1 new or modified test
files. {color} |
|| || || || {color:brown} trunk Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 16m
28s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
28s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
19s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 0m
33s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
11m 14s{color} | {color:green} branch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 0m
42s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
25s{color} | {color:green} trunk passed {color} |
|| || || || {color:brown} Patch Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 0m
29s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
26s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javac {color} | {color:green} 0m
26s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
15s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 0m
30s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} whitespace {color} | {color:green} 0m
0s{color} | {color:green} The patch has no whitespace issues. {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
11m 59s{color} | {color:green} patch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 0m
45s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
17s{color} | {color:green} the patch passed {color} |
|| || || || {color:brown} Other Tests {color} ||
| {color:green}+1{color} | {color:green} unit {color} | {color:green} 4m
32s{color} | {color:green} hadoop-aws in the patch passed. {color} |
| {color:green}+1{color} | {color:green} asflicense {color} | {color:green} 0m
24s{color} | {color:green} The patch does not generate ASF License warnings.
{color} |
| {color:black}{color} | {color:black} {color} | {color:black} 50m 12s{color} |
{color:black} {color} |
\\
\\
|| Subsystem || Report/Notes ||
| Docker | Client=17.05.0-ce Server=17.05.0-ce Image:yetus/hadoop:8f97d6f |
| JIRA Issue | HADOOP-15999 |
| JIRA Patch URL |
https://issues.apache.org/jira/secure/attachment/12959424/HADOOP-15999.003.patch
|
| Optional Tests | dupname asflicense compile javac javadoc mvninstall
mvnsite unit shadedclient findbugs checkstyle |
| uname | Linux defa0984c3df 4.4.0-139-generic #165-Ubuntu SMP Wed Oct 24
10:58:50 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux |
| Build tool | maven |
| Personality | /testptch/patchprocess/precommit/personality/provided.sh |
| git revision | trunk / 41e18fe |
| maven | version: Apache Maven 3.3.9 |
| Default Java | 1.8.0_191 |
| findbugs | v3.1.0-RC1 |
| Test Results |
https://builds.apache.org/job/PreCommit-HADOOP-Build/15949/testReport/ |
| Max. process+thread count | 446 (vs. ulimit of 1) |
| modules | C: hadoop-tools/hadoop-aws U: hadoop-tools/hadoop-aws |
| Console output |
https://builds.apache.org/job/PreCommit-HADOOP-Build/15949/console |
| Powered by | Apache Yetus 0.8.0 http://yetus.apache.org |
This message was automatically generated.
> [s3a] Better support for out-of-band operations
> ---
>
> Key: HADOOP-15999
> URL:
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16772956#comment-16772956 ] Gabor Bota commented on HADOOP-15999: - patch v3: run against Ireland successful added the docs to md fixed style and other minor issues > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: Sub-task > Components: fs/s3 >Affects Versions: 3.1.0 >Reporter: Sean Mackrory >Assignee: Gabor Bota >Priority: Major > Attachments: HADOOP-15999.001.patch, HADOOP-15999.002.patch, > HADOOP-15999.003.patch, out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16771902#comment-16771902 ] Steve Loughran commented on HADOOP-15999: - +1 for merging docs & code. That metrics JIRA is big enough that it should stay separate —so don't worry about it here. in which case, all that should be needed here is: docs, style changes and a rerun of the tests > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: Sub-task > Components: fs/s3 >Affects Versions: 3.1.0 >Reporter: Sean Mackrory >Assignee: Gabor Bota >Priority: Major > Attachments: HADOOP-15999.001.patch, HADOOP-15999.002.patch, > out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16771897#comment-16771897 ] Gabor Bota commented on HADOOP-15999: - Thanks for the review [~ste...@apache.org]! Docs: It's HADOOP-15780, but I can do the docs here and we can resolve that issue separately without a patch. Metrics at S3AFs: HADOOP-15779, but I will do the relevant part in this jira I'll fix all the other issues. > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: Sub-task > Components: fs/s3 >Affects Versions: 3.1.0 >Reporter: Sean Mackrory >Assignee: Gabor Bota >Priority: Major > Attachments: HADOOP-15999.001.patch, HADOOP-15999.002.patch, > out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[
https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16765059#comment-16765059
]
Steve Loughran commented on HADOOP-15999:
-
-1
not quite yet I'm afraid, though we are close. And this reminds me that without
resilience to rename/delete failures auth mode still has issues.
h3. Docs
It's going to need an update to the s3guard.md doc, something about "S3Guard
and out-of-band IO".
h3. S3AFileSystem
I think we should add another metric/counter of S3Guard out-of-band events.
That way changes aren't just silently swallowed. It also allows our unit tests
to call getStorageStatistics to measure what S3AFileSystem has seen.
* L2307 "S3aFileStatus is null" is correct from the view of the code, but
meaningless for users.And it doesn't include the actual path. Better
"Failed to find file {}. Either it is not yet visible, or it has been deleted"
* L2313 nit: Space between if and (
h3. S3Guard.java
L239: put a space between if and ()
L241. Don't call toString() on the status, let SLF4J do it on demand. Also, fix
typo on "wtih"
h3. ITestS3GuardOutOfBandOperations
* check your import ordering
* L46: use a section round the bits of the javadoc you want to stay
formatted all the way through to javadocs. If you are really ambitious: build
the javadocs to see
* L210: you can use ContractTestUtils.writeTextFile() to write that text file.
* L233: overwriteFile shoud use the path() method to return a path for testing.
These are (a) in a subtree unique to each JVM testing in parallel and (b)
cleaned up in test teardown.
I worry about the switching between modes of the filesystem. Would it be better
to create a new FS instance which has been switched to no store, e..g a
guardedFS and a rawFS? That'd make clearer what is happening.
* L312. swap order of assertEquals() args
* L333. again, use path()
* L379 expectExceptionWhenReading(). Use LambdaTestUtils.intercept, and have
the closure actually return the contents of the file
> [s3a] Better support for out-of-band operations
> ---
>
> Key: HADOOP-15999
> URL: https://issues.apache.org/jira/browse/HADOOP-15999
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3
>Affects Versions: 3.1.0
>Reporter: Sean Mackrory
>Assignee: Gabor Bota
>Priority: Major
> Attachments: HADOOP-15999.001.patch, HADOOP-15999.002.patch,
> out-of-band-operations.patch
>
>
> S3Guard was initially done on the premise that a new MetadataStore would be
> the source of truth, and that it wouldn't provide guarantees if updates were
> done without using S3Guard.
> I've been seeing increased demand for better support for scenarios where
> operations are done on the data that can't reasonably be done with S3Guard
> involved. For example:
> * A file is deleted using S3Guard, and replaced by some other tool. S3Guard
> can't tell the difference between the new file and delete / list
> inconsistency and continues to treat the file as deleted.
> * An S3Guard-ed file is overwritten by a longer file by some other tool. When
> reading the file, only the length of the original file is read.
> We could possibly have smarter behavior here by querying both S3 and the
> MetadataStore (even in cases where we may currently only query the
> MetadataStore in getFileStatus) and use whichever one has the higher modified
> time.
> This kills the performance boost we currently get in some workloads with the
> short-circuited getFileStatus, but we could keep it with authoritative mode
> which should give a larger performance boost. At least we'd get more
> correctness without authoritative mode and a clear declaration of when we can
> make the assumptions required to short-circuit the process. If we can't
> consider S3Guard the source of truth, we need to defer to S3 more.
> We'd need to be extra sure of any locality / time zone issues if we start
> relying on mod_time more directly, but currently we're tracking the
> modification time as returned by S3 anyway.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[
https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16763800#comment-16763800
]
Sean Mackrory commented on HADOOP-15999:
+1 on the patch. I filed HADOOP-15619 to make sure we don't lose Steve's
suggestion of lighter weight alternatives to innerGetFileStatus where possible.
{quote}any time we know we don't care 100% on being consistent{quote}
I was talking to someone recently about slimming down some of the checks done
on create, etc. (or at least make them optional). Otherwise the idea of knowing
we might be less consistent than S3 itself scares me more than a little. Feels
like whenever we've done that someone invariably hit the use case we thought
was fine.
> [s3a] Better support for out-of-band operations
> ---
>
> Key: HADOOP-15999
> URL: https://issues.apache.org/jira/browse/HADOOP-15999
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3
>Affects Versions: 3.1.0
>Reporter: Sean Mackrory
>Assignee: Gabor Bota
>Priority: Major
> Attachments: HADOOP-15999.001.patch, HADOOP-15999.002.patch,
> out-of-band-operations.patch
>
>
> S3Guard was initially done on the premise that a new MetadataStore would be
> the source of truth, and that it wouldn't provide guarantees if updates were
> done without using S3Guard.
> I've been seeing increased demand for better support for scenarios where
> operations are done on the data that can't reasonably be done with S3Guard
> involved. For example:
> * A file is deleted using S3Guard, and replaced by some other tool. S3Guard
> can't tell the difference between the new file and delete / list
> inconsistency and continues to treat the file as deleted.
> * An S3Guard-ed file is overwritten by a longer file by some other tool. When
> reading the file, only the length of the original file is read.
> We could possibly have smarter behavior here by querying both S3 and the
> MetadataStore (even in cases where we may currently only query the
> MetadataStore in getFileStatus) and use whichever one has the higher modified
> time.
> This kills the performance boost we currently get in some workloads with the
> short-circuited getFileStatus, but we could keep it with authoritative mode
> which should give a larger performance boost. At least we'd get more
> correctness without authoritative mode and a clear declaration of when we can
> make the assumptions required to short-circuit the process. If we can't
> consider S3Guard the source of truth, we need to defer to S3 more.
> We'd need to be extra sure of any locality / time zone issues if we start
> relying on mod_time more directly, but currently we're tracking the
> modification time as returned by S3 anyway.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[
https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16760547#comment-16760547
]
Hadoop QA commented on HADOOP-15999:
| (/) *{color:green}+1 overall{color}* |
\\
\\
|| Vote || Subsystem || Runtime || Comment ||
| {color:blue}0{color} | {color:blue} reexec {color} | {color:blue} 0m
18s{color} | {color:blue} Docker mode activated. {color} |
|| || || || {color:brown} Prechecks {color} ||
| {color:green}+1{color} | {color:green} @author {color} | {color:green} 0m
0s{color} | {color:green} The patch does not contain any @author tags. {color} |
| {color:green}+1{color} | {color:green} test4tests {color} | {color:green} 0m
0s{color} | {color:green} The patch appears to include 1 new or modified test
files. {color} |
|| || || || {color:brown} trunk Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 17m
33s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
32s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
22s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 0m
35s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
12m 26s{color} | {color:green} branch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 0m
40s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
24s{color} | {color:green} trunk passed {color} |
|| || || || {color:brown} Patch Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 0m
28s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
27s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javac {color} | {color:green} 0m
27s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
17s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 0m
30s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} whitespace {color} | {color:green} 0m
0s{color} | {color:green} The patch has no whitespace issues. {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
12m 50s{color} | {color:green} patch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 0m
45s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
20s{color} | {color:green} the patch passed {color} |
|| || || || {color:brown} Other Tests {color} ||
| {color:green}+1{color} | {color:green} unit {color} | {color:green} 4m
32s{color} | {color:green} hadoop-aws in the patch passed. {color} |
| {color:green}+1{color} | {color:green} asflicense {color} | {color:green} 0m
29s{color} | {color:green} The patch does not generate ASF License warnings.
{color} |
| {color:black}{color} | {color:black} {color} | {color:black} 53m 39s{color} |
{color:black} {color} |
\\
\\
|| Subsystem || Report/Notes ||
| Docker | Client=17.05.0-ce Server=17.05.0-ce Image:yetus/hadoop:8f97d6f |
| JIRA Issue | HADOOP-15999 |
| JIRA Patch URL |
https://issues.apache.org/jira/secure/attachment/12957588/HADOOP-15999.002.patch
|
| Optional Tests | dupname asflicense compile javac javadoc mvninstall
mvnsite unit shadedclient findbugs checkstyle |
| uname | Linux 19d9030e2f3e 3.13.0-153-generic #203-Ubuntu SMP Thu Jun 14
08:52:28 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux |
| Build tool | maven |
| Personality | /testptch/patchprocess/precommit/personality/provided.sh |
| git revision | trunk / 3efa168 |
| maven | version: Apache Maven 3.3.9 |
| Default Java | 1.8.0_191 |
| findbugs | v3.1.0-RC1 |
| Test Results |
https://builds.apache.org/job/PreCommit-HADOOP-Build/15886/testReport/ |
| Max. process+thread count | 321 (vs. ulimit of 1) |
| modules | C: hadoop-tools/hadoop-aws U: hadoop-tools/hadoop-aws |
| Console output |
https://builds.apache.org/job/PreCommit-HADOOP-Build/15886/console |
| Powered by | Apache Yetus 0.8.0 http://yetus.apache.org |
This message was automatically generated.
> [s3a] Better support for out-of-band operations
> ---
>
> Key: HADOOP-15999
> URL:
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16760519#comment-16760519 ] Gabor Bota commented on HADOOP-15999: - [~ste...@apache.org] I think if we commit this the etag can be fixed, or even I can finish HADOOP-15625 if needed - and then it can be used with this. I haven't seen any activity on that jira in the last few months (last patch is 09/Dec/18). If this could optimised further with that feature we can have another issue to do that. > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: Sub-task > Components: fs/s3 >Affects Versions: 3.1.0 >Reporter: Sean Mackrory >Assignee: Gabor Bota >Priority: Major > Attachments: HADOOP-15999.001.patch, HADOOP-15999.002.patch, > out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16760516#comment-16760516 ] Gabor Bota commented on HADOOP-15999: - patch v2 itests run against Ireland successfully. > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: Sub-task > Components: fs/s3 >Affects Versions: 3.1.0 >Reporter: Sean Mackrory >Assignee: Gabor Bota >Priority: Major > Attachments: HADOOP-15999.001.patch, HADOOP-15999.002.patch, > out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16744420#comment-16744420 ] Steve Loughran commented on HADOOP-15999: - Ideally, yes. A key goal of mine is to keep the # of HEAD requests down, so any time we know we don't care 100% on being consistent, we can skip it. FWIW, this is again making me think "what would a better structure of what has grown to be a very large piece of code be". I'm not going to propose any giant refactoring right now, but have a think while you do this about what would be a cleaner model for the internals of the store (the S3 view, the DDB view) and the public filesystem APIs themselves > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: Sub-task > Components: fs/s3 >Affects Versions: 3.1.0 >Reporter: Sean Mackrory >Assignee: Gabor Bota >Priority: Major > Attachments: HADOOP-15999.001.patch, out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16743988#comment-16743988 ] Gabor Bota commented on HADOOP-15999: - [~ste...@apache.org] so the order would be the following: I'll review HADOOP-15229, HADOOP-15625, and refactor and do the required fixes/corrections in my patch. > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: Sub-task > Components: fs/s3 >Affects Versions: 3.1.0 >Reporter: Sean Mackrory >Assignee: Gabor Bota >Priority: Major > Attachments: HADOOP-15999.001.patch, out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16743062#comment-16743062 ] Steve Loughran commented on HADOOP-15999: - its not quite ready yet. That s3 get file status can be lot more minimal if the existence of a file is known in DDB and a check to check its newness is to be executed. We don't need any check for directories here; a simple HEAD/getObject metadata is enough. If there isn't a file, don't bother with HEAD path +"/" or LIST, which could save a few hundred millis. We also want the etag for HADOOP-15625...I'd like to get that patch in before this one, but I'd also like that one to be done after HADOOP-15229. Now, thinking of that file opening, when we are opening a file, we don't need to do a HEAD of that file, because the moment the GET is kicked off (or the SELECT), that will do it...and we already have an invoker in S3aInputStream which knows to retry. And in HADOOP-15625 that first GET will give the etag which is required to be constant on later reads. sowe should distinguish "getFileStatus() as part of the public API" from internal file-existence-checks where we known the immediately subsequent call will be doing the equivalent anyway. > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: Sub-task > Components: fs/s3 >Affects Versions: 3.1.0 >Reporter: Sean Mackrory >Assignee: Gabor Bota >Priority: Major > Attachments: HADOOP-15999.001.patch, out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16742638#comment-16742638 ] Sean Mackrory commented on HADOOP-15999: After a preliminary review I think this looks pretty good. A couple of requests though: * Let's fix the checkstyle issues if you're not already doing that * Can we apply similar logic to when getting directory listings? I think it'll be quick enough to just include as part of this JIRA. I'll do another review when I'm less rushed & tired, but I don't see any other issues with the code right now. Good stuff - thank you. > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: Sub-task > Components: fs/s3 >Affects Versions: 3.1.0 >Reporter: Sean Mackrory >Assignee: Gabor Bota >Priority: Major > Attachments: HADOOP-15999.001.patch, out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[
https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16742373#comment-16742373
]
Hadoop QA commented on HADOOP-15999:
| (/) *{color:green}+1 overall{color}* |
\\
\\
|| Vote || Subsystem || Runtime || Comment ||
| {color:blue}0{color} | {color:blue} reexec {color} | {color:blue} 0m
16s{color} | {color:blue} Docker mode activated. {color} |
|| || || || {color:brown} Prechecks {color} ||
| {color:green}+1{color} | {color:green} @author {color} | {color:green} 0m
0s{color} | {color:green} The patch does not contain any @author tags. {color} |
| {color:green}+1{color} | {color:green} test4tests {color} | {color:green} 0m
0s{color} | {color:green} The patch appears to include 2 new or modified test
files. {color} |
|| || || || {color:brown} trunk Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 21m
2s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
32s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
20s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 0m
40s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
13m 15s{color} | {color:green} branch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 0m
41s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
22s{color} | {color:green} trunk passed {color} |
|| || || || {color:brown} Patch Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 0m
35s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
26s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javac {color} | {color:green} 0m
26s{color} | {color:green} the patch passed {color} |
| {color:orange}-0{color} | {color:orange} checkstyle {color} | {color:orange}
0m 16s{color} | {color:orange} hadoop-tools/hadoop-aws: The patch generated 4
new + 8 unchanged - 0 fixed = 12 total (was 8) {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 0m
31s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} whitespace {color} | {color:green} 0m
0s{color} | {color:green} The patch has no whitespace issues. {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
13m 39s{color} | {color:green} patch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 0m
48s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
20s{color} | {color:green} the patch passed {color} |
|| || || || {color:brown} Other Tests {color} ||
| {color:green}+1{color} | {color:green} unit {color} | {color:green} 4m
33s{color} | {color:green} hadoop-aws in the patch passed. {color} |
| {color:green}+1{color} | {color:green} asflicense {color} | {color:green} 0m
26s{color} | {color:green} The patch does not generate ASF License warnings.
{color} |
| {color:black}{color} | {color:black} {color} | {color:black} 58m 48s{color} |
{color:black} {color} |
\\
\\
|| Subsystem || Report/Notes ||
| Docker | Client=17.05.0-ce Server=17.05.0-ce Image:yetus/hadoop:8f97d6f |
| JIRA Issue | HADOOP-15999 |
| JIRA Patch URL |
https://issues.apache.org/jira/secure/attachment/12954848/HADOOP-15999.001.patch
|
| Optional Tests | dupname asflicense compile javac javadoc mvninstall
mvnsite unit shadedclient findbugs checkstyle |
| uname | Linux fab85c8c9c2b 4.4.0-139-generic #165~14.04.1-Ubuntu SMP Wed Oct
31 10:55:11 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux |
| Build tool | maven |
| Personality | /testptch/patchprocess/precommit/personality/provided.sh |
| git revision | trunk / 3bb745d |
| maven | version: Apache Maven 3.3.9 |
| Default Java | 1.8.0_191 |
| findbugs | v3.1.0-RC1 |
| checkstyle |
https://builds.apache.org/job/PreCommit-HADOOP-Build/15782/artifact/out/diff-checkstyle-hadoop-tools_hadoop-aws.txt
|
| Test Results |
https://builds.apache.org/job/PreCommit-HADOOP-Build/15782/testReport/ |
| Max. process+thread count | 335 (vs. ulimit of 1) |
| modules | C: hadoop-tools/hadoop-aws U: hadoop-tools/hadoop-aws |
| Console output |
https://builds.apache.org/job/PreCommit-HADOOP-Build/15782/console |
| Powered by | Apache Yetus 0.8.0 http://yetus.apache.org |
This
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16734126#comment-16734126 ] Gabor Bota commented on HADOOP-15999: - I'm going to fix geFileStatus: if S3Guard is not in authoritative mode, we should check S3 for filestatus as well. If the metadata from S3 is more recent than what we have on the MS, we should update the MS and return the fresher metadata from S3. I will extend this with inconsistency detection metrics in HADOOP-15779. Also, if we are not running in authoritative mode, all operations will be slower. If MS is authoritative, then we don't have to read S3. Adding more config knobs to this would make this thing over-complicated. There should be an fsck op to sync things (you are right [~ste...@apache.org]), so if a customer runs with S3Guard auth mode and let's say they made some out-of-band operation like a delete without ms, there's a way to sync with the tool. Note: if there's an out-of-band operation right now it's easier to remove the whole ms with prune then handling ms ddb records one by one. > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: Sub-task > Components: fs/s3 >Affects Versions: 3.1.0 >Reporter: Sean Mackrory >Assignee: Gabor Bota >Priority: Major > Attachments: out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16725964#comment-16725964 ] Steve Loughran commented on HADOOP-15999: - I've been aware/worrying about the deleted file problem. and length. At the very least we need an fsck operation which can be kicked off and (throttled) syncs up. This is needed not just for OOB resync, but for recovery from partial failures of rename/delete where things changed and the table hasn't caught up. For delete and even updates, we could use the TTL. Rather than be the source of truth, have it be the source of recent truth, but after, say 60s then outsource the check for deleted file still being missing to the FS *and trusting it*. > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: Sub-task > Components: fs/s3 >Affects Versions: 3.1.0 >Reporter: Sean Mackrory >Assignee: Gabor Bota >Priority: Major > Attachments: out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16719456#comment-16719456 ] Aaron Fabbri commented on HADOOP-15999: --- Was going to link HADOOP-15780 here, but [~gabor.bota] beat me to it. Currently getFileStatus is always short-circuit as you mentioned. We might just want to make that configurable (separate config knob probably). If we are in "check both MS and S3" mode, we probably want a configurable or pluggable conflict policy. The default would probably be to go into a retry loop waiting for both systems (MetadataStore and S3) to agree. After retry policy is exhausted, throw error or continue depending on the conflict policy. Feel feel to ping me for reviews etc. > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: Sub-task > Components: fs/s3 >Affects Versions: 3.1.0 >Reporter: Sean Mackrory >Assignee: Gabor Bota >Priority: Major > Attachments: out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[
https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16719468#comment-16719468
]
Sean Mackrory commented on HADOOP-15999:
{quote}We might just want to make that configurable (separate config knob
probably). If we are in "check both MS and S3" mode, we probably want a
configurable or pluggable conflict policy.{quote}
Yeah - I also considered addressing the out-of-band deletes problem with a
config (or 2) that governs whether we create and / or honor tombstones. But
that's adding exposed complexity and isn't very elegant. If we can relative
easily just start comparing modification times, then we can offer 2 basic modes:
- S3Guard with authoritative mode, in which the MetadataStore is the source of
truth and we can assume All The Things.
- S3Guard without authoritative mode, in which S3 is the source of truth. We
will always be at least as up to date as S3 appears, and will fix list
consistency as long as S3 doesn't give us evidence to the contrary (i.e. older
modification times or the lack of an update entirely).
> [s3a] Better support for out-of-band operations
> ---
>
> Key: HADOOP-15999
> URL: https://issues.apache.org/jira/browse/HADOOP-15999
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3
>Affects Versions: 3.1.0
>Reporter: Sean Mackrory
>Assignee: Gabor Bota
>Priority: Major
> Attachments: out-of-band-operations.patch
>
>
> S3Guard was initially done on the premise that a new MetadataStore would be
> the source of truth, and that it wouldn't provide guarantees if updates were
> done without using S3Guard.
> I've been seeing increased demand for better support for scenarios where
> operations are done on the data that can't reasonably be done with S3Guard
> involved. For example:
> * A file is deleted using S3Guard, and replaced by some other tool. S3Guard
> can't tell the difference between the new file and delete / list
> inconsistency and continues to treat the file as deleted.
> * An S3Guard-ed file is overwritten by a longer file by some other tool. When
> reading the file, only the length of the original file is read.
> We could possibly have smarter behavior here by querying both S3 and the
> MetadataStore (even in cases where we may currently only query the
> MetadataStore in getFileStatus) and use whichever one has the higher modified
> time.
> This kills the performance boost we currently get in some workloads with the
> short-circuited getFileStatus, but we could keep it with authoritative mode
> which should give a larger performance boost. At least we'd get more
> correctness without authoritative mode and a clear declaration of when we can
> make the assumptions required to short-circuit the process. If we can't
> consider S3Guard the source of truth, we need to defer to S3 more.
> We'd need to be extra sure of any locality / time zone issues if we start
> relying on mod_time more directly, but currently we're tracking the
> modification time as returned by S3 anyway.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16718517#comment-16718517 ] Gabor Bota commented on HADOOP-15999: - Thanks [~mackrorysd] for creating this issue. I've added some links that could be related to this. > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: New Feature > Components: fs/s3 >Affects Versions: 3.1.0 >Reporter: Sean Mackrory >Priority: Major > Attachments: out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
[jira] [Commented] (HADOOP-15999) [s3a] Better support for out-of-band operations
[ https://issues.apache.org/jira/browse/HADOOP-15999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16718241#comment-16718241 ] Sean Mackrory commented on HADOOP-15999: Attaching a test that illustrates some of the nuances of the issue when overwriting files. The test of deleted files doesn't fail with S3Guard, which surprises me, as I know this same basic scenario has caused some problems. Need to dig deeper into why this doesn't reproduce that behavior at some point... > [s3a] Better support for out-of-band operations > --- > > Key: HADOOP-15999 > URL: https://issues.apache.org/jira/browse/HADOOP-15999 > Project: Hadoop Common > Issue Type: New Feature >Reporter: Sean Mackrory >Priority: Major > Attachments: out-of-band-operations.patch > > > S3Guard was initially done on the premise that a new MetadataStore would be > the source of truth, and that it wouldn't provide guarantees if updates were > done without using S3Guard. > I've been seeing increased demand for better support for scenarios where > operations are done on the data that can't reasonably be done with S3Guard > involved. For example: > * A file is deleted using S3Guard, and replaced by some other tool. S3Guard > can't tell the difference between the new file and delete / list > inconsistency and continues to treat the file as deleted. > * An S3Guard-ed file is overwritten by a longer file by some other tool. When > reading the file, only the length of the original file is read. > We could possibly have smarter behavior here by querying both S3 and the > MetadataStore (even in cases where we may currently only query the > MetadataStore in getFileStatus) and use whichever one has the higher modified > time. > This kills the performance boost we currently get in some workloads with the > short-circuited getFileStatus, but we could keep it with authoritative mode > which should give a larger performance boost. At least we'd get more > correctness without authoritative mode and a clear declaration of when we can > make the assumptions required to short-circuit the process. If we can't > consider S3Guard the source of truth, we need to defer to S3 more. > We'd need to be extra sure of any locality / time zone issues if we start > relying on mod_time more directly, but currently we're tracking the > modification time as returned by S3 anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
