[jira] [Updated] (HADOOP-15478) WASB: hflush() and hsync() regression
[ https://issues.apache.org/jira/browse/HADOOP-15478?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Thomas Marquardt updated HADOOP-15478: -- Attachment: HADOOP-15547.002.patch > WASB: hflush() and hsync() regression > - > > Key: HADOOP-15478 > URL: https://issues.apache.org/jira/browse/HADOOP-15478 > Project: Hadoop Common > Issue Type: Bug > Components: fs/azure >Affects Versions: 2.9.0, 3.0.2 >Reporter: Thomas Marquardt >Assignee: Thomas Marquardt >Priority: Major > Fix For: 2.10.0, 3.1.1 > > Attachments: HADOOP-15478-002.patch, HADOOP-15478.001.patch > > > HADOOP-14520 introduced a regression in hflush() and hsync(). Previously, > for the default case where users upload data as block blobs, these were > no-ops. Unfortunately, HADOOP-14520 accidentally implemented hflush() and > hsync() by default, so any data buffered in the stream is immediately > uploaded to storage. This new behavior is undesirable, because block blobs > have a limit of 50,000 blocks. Spark users are now seeing failures due to > exceeding the block limit, since Spark frequently invokes hflush(). -- 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] [Updated] (HADOOP-15478) WASB: hflush() and hsync() regression
[ https://issues.apache.org/jira/browse/HADOOP-15478?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Thomas Marquardt updated HADOOP-15478: -- Attachment: (was: HADOOP-15547.002.patch) > WASB: hflush() and hsync() regression > - > > Key: HADOOP-15478 > URL: https://issues.apache.org/jira/browse/HADOOP-15478 > Project: Hadoop Common > Issue Type: Bug > Components: fs/azure >Affects Versions: 2.9.0, 3.0.2 >Reporter: Thomas Marquardt >Assignee: Thomas Marquardt >Priority: Major > Fix For: 2.10.0, 3.1.1 > > Attachments: HADOOP-15478-002.patch, HADOOP-15478.001.patch > > > HADOOP-14520 introduced a regression in hflush() and hsync(). Previously, > for the default case where users upload data as block blobs, these were > no-ops. Unfortunately, HADOOP-14520 accidentally implemented hflush() and > hsync() by default, so any data buffered in the stream is immediately > uploaded to storage. This new behavior is undesirable, because block blobs > have a limit of 50,000 blocks. Spark users are now seeing failures due to > exceeding the block limit, since Spark frequently invokes hflush(). -- 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] [Updated] (HADOOP-15478) WASB: hflush() and hsync() regression
[
https://issues.apache.org/jira/browse/HADOOP-15478?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Steve Loughran updated HADOOP-15478:
Resolution: Fixed
Fix Version/s: 3.1.1
2.10.0
Status: Resolved (was: Patch Available)
+1
applied to branch-3.1&, then cherrypicked to branch-2 and retested against
Azure ireland. new test was happy.
I did see a failure of {{ITestAzureFileSystemInstrumentation}}; its assertion
about bytes written in the last second, even when I tried a standalone run of it
{code}
[ERROR]
ITestAzureFileSystemInstrumentation.testMetricsOnFileCreateRead:162->Assert.assertTrue:41->Assert.fail:88
The bytes written in the last second 0 is pretty far from the expected range
of around 1000 bytes plus a little overhead.
{code}
I think my network is just playing up today, with bandwidth/latency below what
the tests expect. If its recurrent, we might have to think about making the
assertion checks tuneable.
> WASB: hflush() and hsync() regression
> -
>
> Key: HADOOP-15478
> URL: https://issues.apache.org/jira/browse/HADOOP-15478
> Project: Hadoop Common
> Issue Type: Bug
> Components: fs/azure
>Affects Versions: 2.9.0, 3.0.2
>Reporter: Thomas Marquardt
>Assignee: Thomas Marquardt
>Priority: Major
> Fix For: 2.10.0, 3.1.1
>
> Attachments: HADOOP-15478-002.patch, HADOOP-15478.001.patch
>
>
> HADOOP-14520 introduced a regression in hflush() and hsync(). Previously,
> for the default case where users upload data as block blobs, these were
> no-ops. Unfortunately, HADOOP-14520 accidentally implemented hflush() and
> hsync() by default, so any data buffered in the stream is immediately
> uploaded to storage. This new behavior is undesirable, because block blobs
> have a limit of 50,000 blocks. Spark users are now seeing failures due to
> exceeding the block limit, since Spark frequently invokes hflush().
--
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] [Updated] (HADOOP-15478) WASB: hflush() and hsync() regression
[ https://issues.apache.org/jira/browse/HADOOP-15478?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Steve Loughran updated HADOOP-15478: Attachment: HADOOP-15478-002.patch > WASB: hflush() and hsync() regression > - > > Key: HADOOP-15478 > URL: https://issues.apache.org/jira/browse/HADOOP-15478 > Project: Hadoop Common > Issue Type: Bug > Components: fs/azure >Affects Versions: 2.9.0, 3.0.2 >Reporter: Thomas Marquardt >Assignee: Thomas Marquardt >Priority: Major > Attachments: HADOOP-15478-002.patch, HADOOP-15478.001.patch > > > HADOOP-14520 introduced a regression in hflush() and hsync(). Previously, > for the default case where users upload data as block blobs, these were > no-ops. Unfortunately, HADOOP-14520 accidentally implemented hflush() and > hsync() by default, so any data buffered in the stream is immediately > uploaded to storage. This new behavior is undesirable, because block blobs > have a limit of 50,000 blocks. Spark users are now seeing failures due to > exceeding the block limit, since Spark frequently invokes hflush(). -- 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] [Updated] (HADOOP-15478) WASB: hflush() and hsync() regression
[ https://issues.apache.org/jira/browse/HADOOP-15478?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Thomas Marquardt updated HADOOP-15478: -- Status: Patch Available (was: Open) Submitting patch HADOOP-15478.001.patch > WASB: hflush() and hsync() regression > - > > Key: HADOOP-15478 > URL: https://issues.apache.org/jira/browse/HADOOP-15478 > Project: Hadoop Common > Issue Type: Bug > Components: fs/azure >Affects Versions: 3.0.2, 2.9.0 >Reporter: Thomas Marquardt >Assignee: Thomas Marquardt >Priority: Major > Attachments: HADOOP-15478.001.patch > > > HADOOP-14520 introduced a regression in hflush() and hsync(). Previously, > for the default case where users upload data as block blobs, these were > no-ops. Unfortunately, HADOOP-14520 accidentally implemented hflush() and > hsync() by default, so any data buffered in the stream is immediately > uploaded to storage. This new behavior is undesirable, because block blobs > have a limit of 50,000 blocks. Spark users are now seeing failures due to > exceeding the block limit, since Spark frequently invokes hflush(). -- 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] [Updated] (HADOOP-15478) WASB: hflush() and hsync() regression
[ https://issues.apache.org/jira/browse/HADOOP-15478?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Thomas Marquardt updated HADOOP-15478: -- Attachment: HADOOP-15478.001.patch > WASB: hflush() and hsync() regression > - > > Key: HADOOP-15478 > URL: https://issues.apache.org/jira/browse/HADOOP-15478 > Project: Hadoop Common > Issue Type: Bug > Components: fs/azure >Affects Versions: 2.9.0, 3.0.2 >Reporter: Thomas Marquardt >Assignee: Thomas Marquardt >Priority: Major > Attachments: HADOOP-15478.001.patch > > > HADOOP-14520 introduced a regression in hflush() and hsync(). Previously, > for the default case where users upload data as block blobs, these were > no-ops. Unfortunately, HADOOP-14520 accidentally implemented hflush() and > hsync() by default, so any data buffered in the stream is immediately > uploaded to storage. This new behavior is undesirable, because block blobs > have a limit of 50,000 blocks. Spark users are now seeing failures due to > exceeding the block limit, since Spark frequently invokes hflush(). -- 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
