[jira] [Updated] (HDFS-4025) QJM: Sychronize past log segments to JNs that missed them
[ https://issues.apache.org/jira/browse/HDFS-4025?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Jing Zhao updated HDFS-4025: Resolution: Fixed Hadoop Flags: Reviewed Fix Version/s: 3.0.0-alpha3 Status: Resolved (was: Patch Available) I've committed the patch to trunk. Thanks for the contribution Hanisha! > QJM: Sychronize past log segments to JNs that missed them > - > > Key: HDFS-4025 > URL: https://issues.apache.org/jira/browse/HDFS-4025 > Project: Hadoop HDFS > Issue Type: Sub-task > Components: ha >Affects Versions: QuorumJournalManager (HDFS-3077) >Reporter: Todd Lipcon >Assignee: Hanisha Koneru > Fix For: 3.0.0-alpha3, QuorumJournalManager (HDFS-3077) > > Attachments: HDFS-4025.000.patch, HDFS-4025.001.patch, > HDFS-4025.002.patch, HDFS-4025.003.patch, HDFS-4025.004.patch, > HDFS-4025.005.patch, HDFS-4025.006.patch, HDFS-4025.007.patch, > HDFS-4025.008.patch, HDFS-4025.009.patch, HDFS-4025.010.patch, > HDFS-4025.011.patch > > > Currently, if a JournalManager crashes and misses some segment of logs, and > then comes back, it will be re-added as a valid part of the quorum on the > next log roll. However, it will not have a complete history of log segments > (i.e any individual JN may have gaps in its transaction history). This > mirrors the behavior of the NameNode when there are multiple local > directories specified. > However, it would be better if a background thread noticed these gaps and > "filled them in" by grabbing the segments from other JournalNodes. This > increases the resilience of the system when JournalNodes get reformatted or > otherwise lose their local disk. -- This message was sent by Atlassian JIRA (v6.3.15#6346) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-4025) QJM: Sychronize past log segments to JNs that missed them
[ https://issues.apache.org/jira/browse/HDFS-4025?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Hanisha Koneru updated HDFS-4025: - Attachment: HDFS-4025.011.patch Turned sync off by default in patch v11. > QJM: Sychronize past log segments to JNs that missed them > - > > Key: HDFS-4025 > URL: https://issues.apache.org/jira/browse/HDFS-4025 > Project: Hadoop HDFS > Issue Type: Sub-task > Components: ha >Affects Versions: QuorumJournalManager (HDFS-3077) >Reporter: Todd Lipcon >Assignee: Hanisha Koneru > Fix For: QuorumJournalManager (HDFS-3077) > > Attachments: HDFS-4025.000.patch, HDFS-4025.001.patch, > HDFS-4025.002.patch, HDFS-4025.003.patch, HDFS-4025.004.patch, > HDFS-4025.005.patch, HDFS-4025.006.patch, HDFS-4025.007.patch, > HDFS-4025.008.patch, HDFS-4025.009.patch, HDFS-4025.010.patch, > HDFS-4025.011.patch > > > Currently, if a JournalManager crashes and misses some segment of logs, and > then comes back, it will be re-added as a valid part of the quorum on the > next log roll. However, it will not have a complete history of log segments > (i.e any individual JN may have gaps in its transaction history). This > mirrors the behavior of the NameNode when there are multiple local > directories specified. > However, it would be better if a background thread noticed these gaps and > "filled them in" by grabbing the segments from other JournalNodes. This > increases the resilience of the system when JournalNodes get reformatted or > otherwise lose their local disk. -- This message was sent by Atlassian JIRA (v6.3.15#6346) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-4025) QJM: Sychronize past log segments to JNs that missed them
[ https://issues.apache.org/jira/browse/HDFS-4025?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Hanisha Koneru updated HDFS-4025: - Attachment: HDFS-4025.010.patch Thank you [~jingzhao] for reviewing the patch. I have addressed your comments in patch v10. > QJM: Sychronize past log segments to JNs that missed them > - > > Key: HDFS-4025 > URL: https://issues.apache.org/jira/browse/HDFS-4025 > Project: Hadoop HDFS > Issue Type: Sub-task > Components: ha >Affects Versions: QuorumJournalManager (HDFS-3077) >Reporter: Todd Lipcon >Assignee: Hanisha Koneru > Fix For: QuorumJournalManager (HDFS-3077) > > Attachments: HDFS-4025.000.patch, HDFS-4025.001.patch, > HDFS-4025.002.patch, HDFS-4025.003.patch, HDFS-4025.004.patch, > HDFS-4025.005.patch, HDFS-4025.006.patch, HDFS-4025.007.patch, > HDFS-4025.008.patch, HDFS-4025.009.patch, HDFS-4025.010.patch > > > Currently, if a JournalManager crashes and misses some segment of logs, and > then comes back, it will be re-added as a valid part of the quorum on the > next log roll. However, it will not have a complete history of log segments > (i.e any individual JN may have gaps in its transaction history). This > mirrors the behavior of the NameNode when there are multiple local > directories specified. > However, it would be better if a background thread noticed these gaps and > "filled them in" by grabbing the segments from other JournalNodes. This > increases the resilience of the system when JournalNodes get reformatted or > otherwise lose their local disk. -- This message was sent by Atlassian JIRA (v6.3.15#6346) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-4025) QJM: Sychronize past log segments to JNs that missed them
[ https://issues.apache.org/jira/browse/HDFS-4025?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Hanisha Koneru updated HDFS-4025: - Attachment: HDFS-4025.009.patch Patch v09 has some changes. * Checkstyle corrections * Minor optimizations * LastSyncedTxId has been dropped. It's possible that a checkpoint is done during sync interval and some segments are purged. To ensure correct lastSyncedTxId, we would need the checkpointing information from the NameNode. We can skip this doing this for now and always sync from the first transaction id. This improvement can be added later on. Thank you [~jingzhao] for guiding me through with this. > QJM: Sychronize past log segments to JNs that missed them > - > > Key: HDFS-4025 > URL: https://issues.apache.org/jira/browse/HDFS-4025 > Project: Hadoop HDFS > Issue Type: Sub-task > Components: ha >Affects Versions: QuorumJournalManager (HDFS-3077) >Reporter: Todd Lipcon >Assignee: Hanisha Koneru > Fix For: QuorumJournalManager (HDFS-3077) > > Attachments: HDFS-4025.000.patch, HDFS-4025.001.patch, > HDFS-4025.002.patch, HDFS-4025.003.patch, HDFS-4025.004.patch, > HDFS-4025.005.patch, HDFS-4025.006.patch, HDFS-4025.007.patch, > HDFS-4025.008.patch, HDFS-4025.009.patch > > > Currently, if a JournalManager crashes and misses some segment of logs, and > then comes back, it will be re-added as a valid part of the quorum on the > next log roll. However, it will not have a complete history of log segments > (i.e any individual JN may have gaps in its transaction history). This > mirrors the behavior of the NameNode when there are multiple local > directories specified. > However, it would be better if a background thread noticed these gaps and > "filled them in" by grabbing the segments from other JournalNodes. This > increases the resilience of the system when JournalNodes get reformatted or > otherwise lose their local disk. -- This message was sent by Atlassian JIRA (v6.3.15#6346) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-4025) QJM: Sychronize past log segments to JNs that missed them
[ https://issues.apache.org/jira/browse/HDFS-4025?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Hanisha Koneru updated HDFS-4025: - Attachment: HDFS-4025.008.patch Thank you [~jingzhao] for reviewing the patch. {quote} 5. Similarly please see if we still need JNStorage#getTemporaryEditsFile and JNStorage#getFinalizedEditsFile. {quote} We would need these two methods as the corresponding methods in NNStorage require the current storage directory to passed as arguments. {quote} 12. The whole "getMissingLogSegments" may need to be redesigned: Each time we download a missing segment successfully, we should update lastSyncedTxId accordingly. {quote} Suppose the lastSyncedTxId is 10 and the other journal node from which we are downloading missing logs has logs starting from edits_20_30. then we should not update the lastSyncedTxId to 30 as we might still get the missing edits 11 to 20 in another journal node. Instead, if we update the lastSyncedTxId at the end of one sync cycle (after downloading all missing logs from a journal), then we can avoid this scenario. I have addressed rest of the comments in patch v08. > QJM: Sychronize past log segments to JNs that missed them > - > > Key: HDFS-4025 > URL: https://issues.apache.org/jira/browse/HDFS-4025 > Project: Hadoop HDFS > Issue Type: Sub-task > Components: ha >Affects Versions: QuorumJournalManager (HDFS-3077) >Reporter: Todd Lipcon >Assignee: Hanisha Koneru > Fix For: QuorumJournalManager (HDFS-3077) > > Attachments: HDFS-4025.000.patch, HDFS-4025.001.patch, > HDFS-4025.002.patch, HDFS-4025.003.patch, HDFS-4025.004.patch, > HDFS-4025.005.patch, HDFS-4025.006.patch, HDFS-4025.007.patch, > HDFS-4025.008.patch > > > Currently, if a JournalManager crashes and misses some segment of logs, and > then comes back, it will be re-added as a valid part of the quorum on the > next log roll. However, it will not have a complete history of log segments > (i.e any individual JN may have gaps in its transaction history). This > mirrors the behavior of the NameNode when there are multiple local > directories specified. > However, it would be better if a background thread noticed these gaps and > "filled them in" by grabbing the segments from other JournalNodes. This > increases the resilience of the system when JournalNodes get reformatted or > otherwise lose their local disk. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-4025) QJM: Sychronize past log segments to JNs that missed them
[ https://issues.apache.org/jira/browse/HDFS-4025?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Hanisha Koneru updated HDFS-4025: - Attachment: HDFS-4025.007.patch Thank you [~jingzhao] for reviewing the patch and for the comments. I have tried addressing all your comments in patch v07. > QJM: Sychronize past log segments to JNs that missed them > - > > Key: HDFS-4025 > URL: https://issues.apache.org/jira/browse/HDFS-4025 > Project: Hadoop HDFS > Issue Type: Sub-task > Components: ha >Affects Versions: QuorumJournalManager (HDFS-3077) >Reporter: Todd Lipcon >Assignee: Hanisha Koneru > Fix For: QuorumJournalManager (HDFS-3077) > > Attachments: HDFS-4025.000.patch, HDFS-4025.001.patch, > HDFS-4025.002.patch, HDFS-4025.003.patch, HDFS-4025.004.patch, > HDFS-4025.005.patch, HDFS-4025.006.patch, HDFS-4025.007.patch > > > Currently, if a JournalManager crashes and misses some segment of logs, and > then comes back, it will be re-added as a valid part of the quorum on the > next log roll. However, it will not have a complete history of log segments > (i.e any individual JN may have gaps in its transaction history). This > mirrors the behavior of the NameNode when there are multiple local > directories specified. > However, it would be better if a background thread noticed these gaps and > "filled them in" by grabbing the segments from other JournalNodes. This > increases the resilience of the system when JournalNodes get reformatted or > otherwise lose their local disk. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-4025) QJM: Sychronize past log segments to JNs that missed them
[ https://issues.apache.org/jira/browse/HDFS-4025?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Hanisha Koneru updated HDFS-4025: - Attachment: HDFS-4025.006.patch Patch v06 re-uses the #doGetUrl code from TransferFsImage (now in Util). > QJM: Sychronize past log segments to JNs that missed them > - > > Key: HDFS-4025 > URL: https://issues.apache.org/jira/browse/HDFS-4025 > Project: Hadoop HDFS > Issue Type: Sub-task > Components: ha >Affects Versions: QuorumJournalManager (HDFS-3077) >Reporter: Todd Lipcon >Assignee: Hanisha Koneru > Fix For: QuorumJournalManager (HDFS-3077) > > Attachments: HDFS-4025.000.patch, HDFS-4025.001.patch, > HDFS-4025.002.patch, HDFS-4025.003.patch, HDFS-4025.004.patch, > HDFS-4025.005.patch, HDFS-4025.006.patch > > > Currently, if a JournalManager crashes and misses some segment of logs, and > then comes back, it will be re-added as a valid part of the quorum on the > next log roll. However, it will not have a complete history of log segments > (i.e any individual JN may have gaps in its transaction history). This > mirrors the behavior of the NameNode when there are multiple local > directories specified. > However, it would be better if a background thread noticed these gaps and > "filled them in" by grabbing the segments from other JournalNodes. This > increases the resilience of the system when JournalNodes get reformatted or > otherwise lose their local disk. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-4025) QJM: Sychronize past log segments to JNs that missed them
[ https://issues.apache.org/jira/browse/HDFS-4025?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Hanisha Koneru updated HDFS-4025: - Attachment: HDFS-4025.005.patch Fixed checkstyle, findbugs and junit test errors in v5 patch. TestSecureNNWithQJM and TestRollingUpgrade tests fail on trunk as well. > QJM: Sychronize past log segments to JNs that missed them > - > > Key: HDFS-4025 > URL: https://issues.apache.org/jira/browse/HDFS-4025 > Project: Hadoop HDFS > Issue Type: Sub-task > Components: ha >Affects Versions: QuorumJournalManager (HDFS-3077) >Reporter: Todd Lipcon >Assignee: Hanisha Koneru > Fix For: QuorumJournalManager (HDFS-3077) > > Attachments: HDFS-4025.000.patch, HDFS-4025.001.patch, > HDFS-4025.002.patch, HDFS-4025.003.patch, HDFS-4025.004.patch, > HDFS-4025.005.patch > > > Currently, if a JournalManager crashes and misses some segment of logs, and > then comes back, it will be re-added as a valid part of the quorum on the > next log roll. However, it will not have a complete history of log segments > (i.e any individual JN may have gaps in its transaction history). This > mirrors the behavior of the NameNode when there are multiple local > directories specified. > However, it would be better if a background thread noticed these gaps and > "filled them in" by grabbing the segments from other JournalNodes. This > increases the resilience of the system when JournalNodes get reformatted or > otherwise lose their local disk. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-4025) QJM: Sychronize past log segments to JNs that missed them
[ https://issues.apache.org/jira/browse/HDFS-4025?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Hanisha Koneru updated HDFS-4025: - Attachment: HDFS-4025.004.patch Thank you [~jingzhao] for reviewing the patch. I have addressed your comments in v4 patch. Reusing TransferFsImage code would require adding dependencies. > QJM: Sychronize past log segments to JNs that missed them > - > > Key: HDFS-4025 > URL: https://issues.apache.org/jira/browse/HDFS-4025 > Project: Hadoop HDFS > Issue Type: Sub-task > Components: ha >Affects Versions: QuorumJournalManager (HDFS-3077) >Reporter: Todd Lipcon >Assignee: Hanisha Koneru > Fix For: QuorumJournalManager (HDFS-3077) > > Attachments: HDFS-4025.000.patch, HDFS-4025.001.patch, > HDFS-4025.002.patch, HDFS-4025.003.patch, HDFS-4025.004.patch > > > Currently, if a JournalManager crashes and misses some segment of logs, and > then comes back, it will be re-added as a valid part of the quorum on the > next log roll. However, it will not have a complete history of log segments > (i.e any individual JN may have gaps in its transaction history). This > mirrors the behavior of the NameNode when there are multiple local > directories specified. > However, it would be better if a background thread noticed these gaps and > "filled them in" by grabbing the segments from other JournalNodes. This > increases the resilience of the system when JournalNodes get reformatted or > otherwise lose their local disk. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-4025) QJM: Sychronize past log segments to JNs that missed them
[ https://issues.apache.org/jira/browse/HDFS-4025?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Hanisha Koneru updated HDFS-4025: - Attachment: HDFS-4025.003.patch > QJM: Sychronize past log segments to JNs that missed them > - > > Key: HDFS-4025 > URL: https://issues.apache.org/jira/browse/HDFS-4025 > Project: Hadoop HDFS > Issue Type: Sub-task > Components: ha >Affects Versions: QuorumJournalManager (HDFS-3077) >Reporter: Todd Lipcon >Assignee: Hanisha Koneru > Fix For: QuorumJournalManager (HDFS-3077) > > Attachments: HDFS-4025.000.patch, HDFS-4025.001.patch, > HDFS-4025.002.patch, HDFS-4025.003.patch > > > Currently, if a JournalManager crashes and misses some segment of logs, and > then comes back, it will be re-added as a valid part of the quorum on the > next log roll. However, it will not have a complete history of log segments > (i.e any individual JN may have gaps in its transaction history). This > mirrors the behavior of the NameNode when there are multiple local > directories specified. > However, it would be better if a background thread noticed these gaps and > "filled them in" by grabbing the segments from other JournalNodes. This > increases the resilience of the system when JournalNodes get reformatted or > otherwise lose their local disk. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-4025) QJM: Sychronize past log segments to JNs that missed them
[ https://issues.apache.org/jira/browse/HDFS-4025?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Hanisha Koneru updated HDFS-4025: - Attachment: HDFS-4025.002.patch Fixed checkstyle and findbug errors > QJM: Sychronize past log segments to JNs that missed them > - > > Key: HDFS-4025 > URL: https://issues.apache.org/jira/browse/HDFS-4025 > Project: Hadoop HDFS > Issue Type: Sub-task > Components: ha >Affects Versions: QuorumJournalManager (HDFS-3077) >Reporter: Todd Lipcon >Assignee: Hanisha Koneru > Fix For: QuorumJournalManager (HDFS-3077) > > Attachments: HDFS-4025.000.patch, HDFS-4025.001.patch, > HDFS-4025.002.patch > > > Currently, if a JournalManager crashes and misses some segment of logs, and > then comes back, it will be re-added as a valid part of the quorum on the > next log roll. However, it will not have a complete history of log segments > (i.e any individual JN may have gaps in its transaction history). This > mirrors the behavior of the NameNode when there are multiple local > directories specified. > However, it would be better if a background thread noticed these gaps and > "filled them in" by grabbing the segments from other JournalNodes. This > increases the resilience of the system when JournalNodes get reformatted or > otherwise lose their local disk. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-4025) QJM: Sychronize past log segments to JNs that missed them
[ https://issues.apache.org/jira/browse/HDFS-4025?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Hanisha Koneru updated HDFS-4025: - Attachment: HDFS-4025.001.patch Thank you [~jingzhao] for reviewing the patch and suggesting improvements. I have moved al the sync logic to a new class and addressed the other comments as well. > QJM: Sychronize past log segments to JNs that missed them > - > > Key: HDFS-4025 > URL: https://issues.apache.org/jira/browse/HDFS-4025 > Project: Hadoop HDFS > Issue Type: Sub-task > Components: ha >Affects Versions: QuorumJournalManager (HDFS-3077) >Reporter: Todd Lipcon >Assignee: Hanisha Koneru > Fix For: QuorumJournalManager (HDFS-3077) > > Attachments: HDFS-4025.000.patch, HDFS-4025.001.patch > > > Currently, if a JournalManager crashes and misses some segment of logs, and > then comes back, it will be re-added as a valid part of the quorum on the > next log roll. However, it will not have a complete history of log segments > (i.e any individual JN may have gaps in its transaction history). This > mirrors the behavior of the NameNode when there are multiple local > directories specified. > However, it would be better if a background thread noticed these gaps and > "filled them in" by grabbing the segments from other JournalNodes. This > increases the resilience of the system when JournalNodes get reformatted or > otherwise lose their local disk. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-4025) QJM: Sychronize past log segments to JNs that missed them
[ https://issues.apache.org/jira/browse/HDFS-4025?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Jing Zhao updated HDFS-4025: Status: Patch Available (was: Open) > QJM: Sychronize past log segments to JNs that missed them > - > > Key: HDFS-4025 > URL: https://issues.apache.org/jira/browse/HDFS-4025 > Project: Hadoop HDFS > Issue Type: Sub-task > Components: ha >Reporter: Todd Lipcon >Assignee: Todd Lipcon > Attachments: HDFS-4025.000.patch > > > Currently, if a JournalManager crashes and misses some segment of logs, and > then comes back, it will be re-added as a valid part of the quorum on the > next log roll. However, it will not have a complete history of log segments > (i.e any individual JN may have gaps in its transaction history). This > mirrors the behavior of the NameNode when there are multiple local > directories specified. > However, it would be better if a background thread noticed these gaps and > "filled them in" by grabbing the segments from other JournalNodes. This > increases the resilience of the system when JournalNodes get reformatted or > otherwise lose their local disk. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Updated] (HDFS-4025) QJM: Sychronize past log segments to JNs that missed them
[ https://issues.apache.org/jira/browse/HDFS-4025?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Hanisha Koneru updated HDFS-4025: - Attachment: HDFS-4025.000.patch This patch would start a syncJournal thread in each JournalNode. This thread periodically syncs the journal node by getting the edit log manifest from another randomly picked journal node. If the other journal node has an edit log which is missing from the current journal node, it is downloaded from the other journal node's http server. The in progress log segments are not synced in this process. Thank you [~jingzhao] for all the guidance. > QJM: Sychronize past log segments to JNs that missed them > - > > Key: HDFS-4025 > URL: https://issues.apache.org/jira/browse/HDFS-4025 > Project: Hadoop HDFS > Issue Type: Sub-task > Components: ha >Affects Versions: QuorumJournalManager (HDFS-3077) >Reporter: Todd Lipcon >Assignee: Todd Lipcon > Fix For: QuorumJournalManager (HDFS-3077) > > Attachments: HDFS-4025.000.patch > > > Currently, if a JournalManager crashes and misses some segment of logs, and > then comes back, it will be re-added as a valid part of the quorum on the > next log roll. However, it will not have a complete history of log segments > (i.e any individual JN may have gaps in its transaction history). This > mirrors the behavior of the NameNode when there are multiple local > directories specified. > However, it would be better if a background thread noticed these gaps and > "filled them in" by grabbing the segments from other JournalNodes. This > increases the resilience of the system when JournalNodes get reformatted or > otherwise lose their local disk. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org