Reproduce steps:
1. Start SC-1 2. Commit some CCBs # immcfg -c Test test=0 # immcfg -c Test test=1 # immcfg -c Test test=2 # immcfg -c Test test=3 3. Start PL-3 4. Restart SC-1 5. When SC-1 is back, it fails to add operations to CCB # immcfg -c Test test=10 error - saImmOmCcbObjectCreate_2 FAILED with SA_AIS_ERR_FAILED_OPERATION (21) OI reports: IMM: Resource abort: CCB is not in an expected state error - saImmOmCcbApply FAILED: SA_AIS_ERR_FAILED_OPERATION (21) OI reports: IMM: Resource abort: CCB is not in an expected state
cb->mLatestCcbId was not updated on PL-3 when it joined the cluster so it still had value of zero.
When SC-1 was back from headless, IMMND on PL-3 sent re-introduce message to IMMD on SC-1 with cb->mLatestCcbId = 0.
IMMD failed to update cb->ccb_id_count so when new CCB is created, it will start from 0+1 instead of mLatestCcbId + 1.
That results in the conflict with the CCB in sCcbVector and the CCB operation failure.
Attached is logs and traces.
Added "headless" clarification because "SC absence" can be missunderstood as just one (out of normally two) SCs being absent.
Last edit: Anders Bjornerstedt 2017-03-01
default (5.2) [staging:8cbe1c]
changeset: 8669:8cbe1c0a3335
user: Hung Nguyen hung.d.nguyen@dektech.com.au
date: Wed Mar 08 13:14:20 2017 +0700
summary: imm: Sync latest ccb-id to sync clients [#2323]
opensaf-5.1.x [staging:b642a1]
changeset: 8670:b642a1336cf3
user: Hung Nguyen hung.d.nguyen@dektech.com.au
date: Wed Mar 08 13:35:24 2017 +0700
summary: imm: Sync latest ccb-id to sync clients [#2323]
opensaf-5.0.x [staging:bf11d1]
changeset: 8671:bf11d1e7720a
user: Hung Nguyen hung.d.nguyen@dektech.com.au
date: Wed Mar 08 13:35:24 2017 +0700
summary: imm: Sync latest ccb-id to sync clients [#2323]
Related
Commit: [8cbe1c]
Commit: [b642a1]
Commit: [bf11d1]
Tickets:
#2323