-
Notifications
You must be signed in to change notification settings - Fork 3.7k
[fix](fe) Fix drop frontend cause bdbje and fe memory inconsistent state
#35203
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
|
Thank you for your contribution to Apache Doris. Since 2024-03-18, the Document has been moved to doris-website. |
|
run buildall |
TPC-H: Total hot run time: 40883 ms |
TPC-DS: Total hot run time: 169083 ms |
ClickBench: Total hot run time: 30.58 s |
b30a165 to
69362db
Compare
|
run buildall |
TPC-H: Total hot run time: 39799 ms |
TPC-DS: Total hot run time: 172109 ms |
ClickBench: Total hot run time: 30.07 s |
c70f678 to
4994fbd
Compare
|
run buildall |
TPC-H: Total hot run time: 41840 ms |
TPC-DS: Total hot run time: 168818 ms |
ClickBench: Total hot run time: 30.57 s |
…t state * dropFrontend use `ReplicationGroupAdmin` to remove bdbje node, in some corner case `ReplicationGroupAdmin.removeMember` success but `editLog.logRemoveFrontend` fail will cause inconsistent node info between `fe memory` and `bdbje`
4994fbd to
28eac70
Compare
|
run buildall |
TPC-H: Total hot run time: 39666 ms |
TPC-DS: Total hot run time: 171461 ms |
ClickBench: Total hot run time: 29.81 s |
|
PR approved by anyone and no changes requested. |
dataroaring
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
|
PR approved by at least one committer and no changes requested. |
An exception introduced by #35203 cause the frontend start failed In this cases: 1. all frontends are down and recover frontend in different host by the original metadata 2. or the only frontend's ip changed, and start with `-r` option to recover from original metadata
…t state (#35203) * dropFrontend use `ReplicationGroupAdmin` to remove bdbje node, in some corner case `ReplicationGroupAdmin.removeMember` success but `editLog.logRemoveFrontend` fail will cause inconsistent node info between `fe memory` and `bdbje`
An exception introduced by #35203 cause the frontend start failed In this cases: 1. all frontends are down and recover frontend in different host by the original metadata 2. or the only frontend's ip changed, and start with `-r` option to recover from original metadata
```
com.sleepycat.je.rep.MasterStateException: (JE 18.3.12) Cannot remove an active master
at com.sleepycat.je.rep.util.ReplicationGroupAdmin.checkMember(ReplicationGroupAdmin.java:591)
at com.sleepycat.je.rep.util.ReplicationGroupAdmin.removeMember(ReplicationGroupAdmin.java:315)
at org.apache.doris.ha.BDBHA.removeDroppedMember(BDBHA.java:262)
at org.apache.doris.catalog.Env.removeDroppedFrontends(Env.java:3077)
at org.apache.doris.catalog.Env.transferToMaster(Env.java:1520)
at org.apache.doris.catalog.Env.access$1300(Env.java:338)
at org.apache.doris.catalog.Env$5.runOneCycle(Env.java:2792)
at org.apache.doris.common.util.Daemon.run(Daemon.java:116)
```
* the bug is introduced by apache#35203
```
com.sleepycat.je.rep.MasterStateException: (JE 18.3.12) Cannot remove an active master
at com.sleepycat.je.rep.util.ReplicationGroupAdmin.checkMember(ReplicationGroupAdmin.java:591)
at com.sleepycat.je.rep.util.ReplicationGroupAdmin.removeMember(ReplicationGroupAdmin.java:315)
at org.apache.doris.ha.BDBHA.removeDroppedMember(BDBHA.java:262)
at org.apache.doris.catalog.Env.removeDroppedFrontends(Env.java:3077)
at org.apache.doris.catalog.Env.transferToMaster(Env.java:1520)
at org.apache.doris.catalog.Env.access$1300(Env.java:338)
at org.apache.doris.catalog.Env$5.runOneCycle(Env.java:2792)
at org.apache.doris.common.util.Daemon.run(Daemon.java:116)
```
* the bug is introduced by #35203
```
com.sleepycat.je.rep.MasterStateException: (JE 18.3.12) Cannot remove an active master
at com.sleepycat.je.rep.util.ReplicationGroupAdmin.checkMember(ReplicationGroupAdmin.java:591)
at com.sleepycat.je.rep.util.ReplicationGroupAdmin.removeMember(ReplicationGroupAdmin.java:315)
at org.apache.doris.ha.BDBHA.removeDroppedMember(BDBHA.java:262)
at org.apache.doris.catalog.Env.removeDroppedFrontends(Env.java:3077)
at org.apache.doris.catalog.Env.transferToMaster(Env.java:1520)
at org.apache.doris.catalog.Env.access$1300(Env.java:338)
at org.apache.doris.catalog.Env$5.runOneCycle(Env.java:2792)
at org.apache.doris.common.util.Daemon.run(Daemon.java:116)
```
* the bug is introduced by #35203
ReplicationGroupAdminto remove bdbje node, in some corner caseReplicationGroupAdmin.removeMembersuccess buteditLog.logRemoveFrontendfailwill cause inconsistent node info between
fe memoryandbdbje