VPLEX: Distributed Virtual Volumes Service Status Shows as Disconnected
摘要: After an upgrade to GeoSynchrony 6.0 Service Pack 1 Patch 2 (6.0.1.02.00.03), distributed virtual volumes have a service status of "disconnected." This does not affect I/O, and there should be no data unavailability (DU) due to this issue. ...
症状
Impacted VPLEX GeoSynchony code versions:
Software: GeoSynchony 6.0 Service Pack 1 Patch 2
Impacted VPLEX Hardware:
VPLEX Series
VPLEX VS2
VPLEX VS6
VPLEX-Metro
After an upgrade to GeoSynchrony 6.0 Service Pack 1 Patch 2 (6.0.1.02.00.03), distributed virtual volumes have a service status of "disconnected." This does not affect I/O, and the customer should not see DU due to this issue.
The AMF_REEXPORT_QD flag is a newly introduced flag in the firmware. It helps internally keep track of outstanding exports in the case that the AMG master goes down. It is implemented in amf_flags, which is a field that is visible to the CLI.
When the second upgraders come up during NDU, this flag is set as a part of normal operation. Typically, the flag would be unset during the re-export of the AMF. However, when booting, the first re-export skips this phase. This causes the flag to remain set until the next re-export. The flag itself is benign and does not interfere with I/O or system configuration.
When virtual-volume summary is run, the volumes have a service status of "disconnected":
Virtual-volume health summary virtual-volume name health state operational status service status --------------------------------- ------------ ------------------ -------------- <vvol name> ok ok disconnected
The underlying distributed device can also show as disconnected or running.
Example of a distributed device showing as "disconnected":
/distributed-storage/distributed-devices/<distributed-device name>: Attributes: Name Value ---------------------- -------------------------------- application-consistent false auto-resume true block-count 536870912 block-size 4K capacity 2T clusters-involved [cluster-1, cluster-2] geometry raid-1 health-indications [] health-state ok locality distributed<<-- operational-status ok rebuild-allowed true rebuild-eta - rebuild-progress - rebuild-status done rebuild-type full rule-set-name cluster-1-detaches service-status -->>disconnected<<-- stripe-depth - system-id <name> thin-capable true transfer-size 128K virtual-volume <vvol name>
Example of a distributed device showing as "running":
/distributed-storage/distributed-devices/<distributed-device name>: Attributes: Name Value ---------------------- ------------------------------------------ application-consistent false auto-resume true block-count 536874240 block-size 4K capacity 2T clusters-involved consistency-group geometry raid-1 health-indications [] health-state ok locality distributed operational-status ok rebuild-allowed true rebuild-eta - rebuild-progress - rebuild-status done rebuild-type full rule-set-name cluster-2-detaches service-status running <<<< storage-array-family symmetrix stripe-depth - system-id <name> thin-capable true transfer-size 128K virtual-volume <vvol name>
原因
A bug in the server-side FLL code is causing the reexport-qd flag to be read from the AMF. The code actually intends to read the non-splitting-fail-all flag. This causes the CLI to report the volume as disconnected when querying with FLL.
FLL is a feature introduced in GeoSynchrony 6.0 Service Pack 1 Patch 2 (6.0.1.02.00.03). In this feature, some management server tasks are offloaded onto the directors to alleviate possible performance impact to the management server.
解决方案
If this issue is observed after an upgrade to GeoSynchrony 6.0 Service Pack 1 Patch 2 (6.0.1.02.00.03), contact Dell Customer Support and quote this Knowledge Base article ID.