[jira] [Updated] (YARN-3705) forcemanual transitionToStandby in RM-HA automatic-failover mode should change elector state
[
https://issues.apache.org/jira/browse/YARN-3705?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Masatake Iwasaki updated YARN-3705:
---
Attachment: YARN-3705.006.patch
bq. If we call resetLeaderElection inside the rmadmin.transitionToStandby(), it
will cause a infinite loop.
You are right. I need to make sure that resetLeaderElection is not called when
EmbeddedElectorService#becomeStandby calls transitionToStandy. Thanks for the
good catch, [~xgong].
I attached 006. Though I checked that the loop is not caused by starting RM-HA
manually with patched jar, it is difficult to test that in unit test.
forcemanual transitionToStandby in RM-HA automatic-failover mode should
change elector state
Key: YARN-3705
URL: https://issues.apache.org/jira/browse/YARN-3705
Project: Hadoop YARN
Issue Type: Sub-task
Reporter: Masatake Iwasaki
Assignee: Masatake Iwasaki
Attachments: YARN-3705.001.patch, YARN-3705.002.patch,
YARN-3705.003.patch, YARN-3705.004.patch, YARN-3705.005.patch,
YARN-3705.006.patch
Executing {{rmadmin -transitionToStandby --forcemanual}} in
automatic-failover.enabled mode makes ResouceManager standby while keeping
the state of ActiveStandbyElector. It should make elector to quit and rejoin
in order to enable other candidates to promote, otherwise forcemanual
transition should not be allowed in automatic-failover mode in order to avoid
confusion.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Updated] (YARN-3705) forcemanual transitionToStandby in RM-HA automatic-failover mode should change elector state
[
https://issues.apache.org/jira/browse/YARN-3705?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Masatake Iwasaki updated YARN-3705:
---
Attachment: YARN-3705.005.patch
findbugs warnings is about synchronization of {{autoFailoverEnabled}} variable.
{noformat}
M M IS: Inconsistent synchronization of
org.apache.hadoop.yarn.server.resourcemanager.AdminService.autoFailoverEnabled;
locked 66% of time Unsynchronized access at AdminService.java:[line 129]
{noformat}
I removed the checking this variable in 005 because equivalent checking is done
in AdminService#resetLeaderElection.
forcemanual transitionToStandby in RM-HA automatic-failover mode should
change elector state
Key: YARN-3705
URL: https://issues.apache.org/jira/browse/YARN-3705
Project: Hadoop YARN
Issue Type: Sub-task
Reporter: Masatake Iwasaki
Assignee: Masatake Iwasaki
Attachments: YARN-3705.001.patch, YARN-3705.002.patch,
YARN-3705.003.patch, YARN-3705.004.patch, YARN-3705.005.patch
Executing {{rmadmin -transitionToStandby --forcemanual}} in
automatic-failover.enabled mode makes ResouceManager standby while keeping
the state of ActiveStandbyElector. It should make elector to quit and rejoin
in order to enable other candidates to promote, otherwise forcemanual
transition should not be allowed in automatic-failover mode in order to avoid
confusion.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Updated] (YARN-3705) forcemanual transitionToStandby in RM-HA automatic-failover mode should change elector state
[
https://issues.apache.org/jira/browse/YARN-3705?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Masatake Iwasaki updated YARN-3705:
---
Attachment: YARN-3705.004.patch
I should not have used ResourceManager#handleTransitionToStandBy due to the way
of error handling.
{code}
} catch (Exception e) {
LOG.fatal(Failed to transition RM to Standby mode.);
ExitUtil.terminate(1, e);
{code}
004 is just using AdminService#resetLeaderElection. TestWorkPreservingRMRestart
and TestRMRestart are succeeded on my local environment.
forcemanual transitionToStandby in RM-HA automatic-failover mode should
change elector state
Key: YARN-3705
URL: https://issues.apache.org/jira/browse/YARN-3705
Project: Hadoop YARN
Issue Type: Sub-task
Reporter: Masatake Iwasaki
Assignee: Masatake Iwasaki
Attachments: YARN-3705.001.patch, YARN-3705.002.patch,
YARN-3705.003.patch, YARN-3705.004.patch
Executing {{rmadmin -transitionToStandby --forcemanual}} in
automatic-failover.enabled mode makes ResouceManager standby while keeping
the state of ActiveStandbyElector. It should make elector to quit and rejoin
in order to enable other candidates to promote, otherwise forcemanual
transition should not be allowed in automatic-failover mode in order to avoid
confusion.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Updated] (YARN-3705) forcemanual transitionToStandby in RM-HA automatic-failover mode should change elector state
[
https://issues.apache.org/jira/browse/YARN-3705?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Masatake Iwasaki updated YARN-3705:
---
Attachment: YARN-3705.002.patch
I'm attached 002 addressing whitespace warnings. TestWorkPreservingRMRestart is
not related to the code path the patch fixes.
forcemanual transitionToStandby in RM-HA automatic-failover mode should
change elector state
Key: YARN-3705
URL: https://issues.apache.org/jira/browse/YARN-3705
Project: Hadoop YARN
Issue Type: Sub-task
Reporter: Masatake Iwasaki
Assignee: Masatake Iwasaki
Attachments: YARN-3705.001.patch, YARN-3705.002.patch
Executing {{rmadmin -transitionToStandby --forcemanual}} in
automatic-failover.enabled mode makes ResouceManager standby while keeping
the state of ActiveStandbyElector. It should make elector to quit and rejoin
in order to enable other candidates to promote, otherwise forcemanual
transition should not be allowed in automatic-failover mode in order to avoid
confusion.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Updated] (YARN-3705) forcemanual transitionToStandby in RM-HA automatic-failover mode should change elector state
[
https://issues.apache.org/jira/browse/YARN-3705?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Masatake Iwasaki updated YARN-3705:
---
Attachment: YARN-3705.003.patch
The test failure is relevant. ResourceManager#handleTransitionToStandBy is
expected to be used only when automatic failover enabled. I am attaching 003
addressing non automatic failover case too.
forcemanual transitionToStandby in RM-HA automatic-failover mode should
change elector state
Key: YARN-3705
URL: https://issues.apache.org/jira/browse/YARN-3705
Project: Hadoop YARN
Issue Type: Sub-task
Reporter: Masatake Iwasaki
Assignee: Masatake Iwasaki
Attachments: YARN-3705.001.patch, YARN-3705.002.patch,
YARN-3705.003.patch
Executing {{rmadmin -transitionToStandby --forcemanual}} in
automatic-failover.enabled mode makes ResouceManager standby while keeping
the state of ActiveStandbyElector. It should make elector to quit and rejoin
in order to enable other candidates to promote, otherwise forcemanual
transition should not be allowed in automatic-failover mode in order to avoid
confusion.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Updated] (YARN-3705) forcemanual transitionToStandby in RM-HA automatic-failover mode should change elector state
[
https://issues.apache.org/jira/browse/YARN-3705?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Masatake Iwasaki updated YARN-3705:
---
Attachment: YARN-3705.001.patch
I attached 001 which apply quick fix of the issue. I need more consideration to
add monitoring by HealthMonitor to AdminService. If it turns out to be worth
adding, I will file a follow-up JIRA.
forcemanual transitionToStandby in RM-HA automatic-failover mode should
change elector state
Key: YARN-3705
URL: https://issues.apache.org/jira/browse/YARN-3705
Project: Hadoop YARN
Issue Type: Sub-task
Reporter: Masatake Iwasaki
Assignee: Masatake Iwasaki
Attachments: YARN-3705.001.patch
Executing {{rmadmin -transitionToStandby --forcemanual}} in
automatic-failover.enabled mode makes ResouceManager standby while keeping
the state of ActiveStandbyElector. It should make elector to quit and rejoin
in order to enable other candidates to promote, otherwise forcemanual
transition should not be allowed in automatic-failover mode in order to avoid
confusion.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Updated] (YARN-3705) forcemanual transitionToStandby in RM-HA automatic-failover mode should change elector state
[
https://issues.apache.org/jira/browse/YARN-3705?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Masatake Iwasaki updated YARN-3705:
---
Description: Executing {{rmadmin -transitionToStandby --forcemanual}} in
automatic-failover.enabled mode makes ResouceManager standby while keeping the
state of ActiveStandbyElector. It should make elector to quit and rejoin in
order to enable other candidates to promote, otherwise forcemanual transition
should not be allowed in automatic-failover mode in order to avoid confusion.
(was: Executing {{rmadmin -transitionToActive --forcemanual}} and {{rmadmin
-transitionToActive --forcemanual}} in automatic-failover.enabled mode changes
the active/standby state of ResouceManager while keeping the state of
ActiveStandbyElector. It should make elector to quit and rejoin otherwise
forcemanual transition should not be allowed in automatic-failover mode in
order to avoid confusion.)
forcemanual transitionToStandby in RM-HA automatic-failover mode should
change elector state
Key: YARN-3705
URL: https://issues.apache.org/jira/browse/YARN-3705
Project: Hadoop YARN
Issue Type: Sub-task
Reporter: Masatake Iwasaki
Assignee: Masatake Iwasaki
Executing {{rmadmin -transitionToStandby --forcemanual}} in
automatic-failover.enabled mode makes ResouceManager standby while keeping
the state of ActiveStandbyElector. It should make elector to quit and rejoin
in order to enable other candidates to promote, otherwise forcemanual
transition should not be allowed in automatic-failover mode in order to avoid
confusion.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Updated] (YARN-3705) forcemanual transitionToStandby in RM-HA automatic-failover mode should change elector state
[
https://issues.apache.org/jira/browse/YARN-3705?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Masatake Iwasaki updated YARN-3705:
---
Summary: forcemanual transitionToStandby in RM-HA automatic-failover mode
should change elector state (was: forcemanual transition of RM active/standby
state in automatic-failover mode should change elector state)
forcemanual transitionToStandby in RM-HA automatic-failover mode should
change elector state
Key: YARN-3705
URL: https://issues.apache.org/jira/browse/YARN-3705
Project: Hadoop YARN
Issue Type: Sub-task
Reporter: Masatake Iwasaki
Assignee: Masatake Iwasaki
Executing {{rmadmin -transitionToActive --forcemanual}} and {{rmadmin
-transitionToActive --forcemanual}} in automatic-failover.enabled mode
changes the active/standby state of ResouceManager while keeping the state of
ActiveStandbyElector. It should make elector to quit and rejoin otherwise
forcemanual transition should not be allowed in automatic-failover mode in
order to avoid confusion.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
