[
https://issues.apache.org/jira/browse/HBASE-26897?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
May updated HBASE-26897:
------------------------
Description:
We got a cluster of 2 master nodes (hm1, hm2) and 3 slaves nodes (rs1, rs2, rs3)
# hm1 becomes active
# rs3 becomes the meta RS
# rs3 crashes
# hm1 crashes. At this time, the meta-region-server in ZooKeeper is still rs3
# hm2 becomes active, but stucks in "master.HMaster: hbase:meta,,1.1588230740
is NOT online; state=\{1588230740 state=OPEN, ts=1648007404659,
server=rs3,16020,1648007272771}; ServerCrashProcedures=true. Master startup
cannot progress, in holding-pattern until region onlined."
was:
We got a cluster of 2 master nodes (hm1, hm2) and 3 slaves nodes (rs1, rs2, rs3)
# hm1 becomes active
# rs3 becomes the meta RS
# rs3 crashes
# rs2 crashes
# hm1 crashes when handling the crash of rs3. At this time, the
meta-region-server in ZooKeeper is still rs3
# hm2 becomes active, but stucks in "master.HMaster: hbase:meta,,1.1588230740
is NOT online; state=\{1588230740 state=OPEN, ts=1648007404659,
server=rs3,16020,1648007272771}; ServerCrashProcedures=true. Master startup
cannot progress, in holding-pattern until region onlined."
> Crash Meta-RS and active HMaster successively make cluster out of service
> -------------------------------------------------------------------------
>
> Key: HBASE-26897
> URL: https://issues.apache.org/jira/browse/HBASE-26897
> Project: HBase
> Issue Type: Bug
> Affects Versions: 2.4.8
> Reporter: May
> Priority: Major
>
> We got a cluster of 2 master nodes (hm1, hm2) and 3 slaves nodes (rs1, rs2,
> rs3)
>
> # hm1 becomes active
> # rs3 becomes the meta RS
> # rs3 crashes
> # hm1 crashes. At this time, the meta-region-server in ZooKeeper is still rs3
> # hm2 becomes active, but stucks in "master.HMaster:
> hbase:meta,,1.1588230740 is NOT online; state=\{1588230740 state=OPEN,
> ts=1648007404659, server=rs3,16020,1648007272771};
> ServerCrashProcedures=true. Master startup cannot progress, in
> holding-pattern until region onlined."
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
