Menu

#3219 imm: immnd crash in multi partitioned clusters rejoin

5.20.11
fixed
None
defect
imm
nd
minor
False
2020-09-18
2020-09-16
Thuan Tran
No

Under scenario multi partitioned cluster rejoin, IMMND crash as following:

2020-09-15 18:39:59.310 SC-6 osafimmnd[195]: NO Re-introduce-me highestProcessed:4358 highestReceived:4358 ex_immd_node_id=2070f
2020-09-15 18:39:59.310 SC-6 osafimmnd[195]: src/imm/immnd/immnd_evt.c:10158: immnd_evt_proc_pbe_prto_purge_mutations: Assertion 'cb->mRulingEpoch <= evt->info.ctrl.rulingEpoch' failed.

Then node don't reboot as expected even it used to on different partition with current coordinator.

Related

Wiki: ChangeLog-5.20.11

Discussion

  • Thuan Tran

    Thuan Tran - 2020-09-16
    • status: assigned --> review
     
  • Thuan Tran

    Thuan Tran - 2020-09-18
    • status: review --> fixed
     
  • Thuan Tran

    Thuan Tran - 2020-09-18

    commit d2bbaba8c28f68f5cb1a5e620022d673cc03600e
    Author: thuan.tran thuan.tran@dektech.com.au
    Date: Wed Sep 16 14:18:33 2020 +0700

    imm: fix immnd crash in multi partitioned clusters rejoin [#3219]
    
    Before IMMND get re-intro response from IMMD, it may get broadcast
    event from IMMD (e.g: IMMND_EVT_D2ND_PBE_PRTO_PURGE_MUTATIONS) then
    crash because it used to be on different partition. As IMMND crash
    then it cannot reboot node as expected. Solution:
    - IMMND prioritize re-introduce response msg from IMMD.
    - IMMND ignore broadcast events from IMMD if re-introduce on-going.
    
     

Log in to post a comment.