Symptoms
ScenarioEach SDS' DRL is discarded upon upgrade, so a full rebuild is needed upon exiting maintenance mode.
SymptomsUpgrading from 3.5, you will see the entire capacity of the SDS in rebuild when it exits maintenance mode.
This can also happen on other cases when SDS exits IMM.
The log should contain the following:
trc.u.3.8:24/02 03:22:27.970864 0x7f24d1ab2db0:mgPhyDev_InitInternal:03265: Unable to find T-sync bitmap in shared memory, set the entire REBUILD DRL for devId: bfe6642f00070000 device name: /dev/sdb
trc.u.3.8:24/02 03:22:28.012700 0x7f24d1ba2db0:mgPhyDev_InitInternal:03265: Unable to find T-sync bitmap in shared memory, set the entire REBUILD DRL for devId: bfe6643800070009 device name: /dev/sdk
trc.u.3.8:24/02 03:22:28.075503 0x7f24d1b57db0:mgPhyDev_InitInternal:03265: Unable to find T-sync bitmap in shared memory, set the entire REBUILD DRL for devId: bfe6643500070006 device name: /dev/sdh
Impact
IMM exit takes a long time to complete. For upgrades - it may take ~15 hours or larger clusters.
Cause
SDS DRL is accidentally set to volatile during an upgrade.
The SDS' RPM upgrade wipes the shared memory, and so the entire node's data is considered out of sync.
Full rebuild occurs, and the IM/gateway moves to the next SDS.
Resolution
Workaround
N/A
Impacted Versions
3.5.x.x
Fixed In Version
3.5.1.2
3.6
