Unsolved
This post is more than 5 years old
2 Posts
1
1001
Vplex Metro Witness test plan: Two disabled SAN switche on the same site
Hello,
I followed a test plan to validate fonctionnalities of Vplex Metro Witness
Action:
Disable the two Brocade swiche on the primary site.
Verification Status:
the distributed device on the Consistency group not accessible on the secondary site.
Workaround:
- why witness dont work properly? I dont knows
- resume-at-loser commandes have no effet
- Power down Management Server, about 10 minutes after , have no effet.
Question:
Have any action or commande to force the distributed device be accessible on the secondary site.
Regards
Paul
TadhgConcannon
1 Rookie
1 Rookie
•
22 Posts
0
January 19th, 2015 02:00
Hi Paul
Based on my findings on something similar
Each VPLEX Cluster with two or more engines uses a pair of dedicated Fibre Channel switches for intra-cluster communication between the directors within the cluster.
Two redundant Fibre Channel fabrics are created with each switch serving a different fabric.
The loss of a single Fibre Channel switch results in no loss of processing or service.
However If the power supply is pulled from the both FC Switch1 and FC Switch 2 in Cluster1 (same as you both SAN switches are unavailable )
Then the following happens
The directors remain online but lose the ability to communicate with each other
Similar to what you have observed
Witness sees the directors are still online so witness does not intervene - hence witness believes that the directors are still OK and will allow preference rules to fire.
In our configuration the preference rules for 3 consistency groups which were set to site A wins
All preference rules fired, i.e. service was suspended on Cluster-2 for all consistency groups set to “Cluster-1 wins”, with Cluster-1 supposedly continuing service.
However, as the directors are not in communication with each other, the devices / volumes were no longer available on Cluster-1, either.
Result was a full DU situation on Site A
However the storage in Site B that was in consistency group also became unavailable
Storage that were in CGs set to “Cluster-2 wins” also become unusable on Cluster-2 -- this is because the storage on Site A is is unavailable.
Only storage that was local to Site B remained online and useable.
What I believe is required is a fix from engineering so that witness also monitors the status of the SAN switches.
However this is a very unusual scenario in that loosing both SAN switches, but the rest of the VPLEX environment is working correctly would be a very rare event and is not something that would happen normally.
Looking forward to seeing if there is more information available on this
Regards,
Tadhg
Paul.Marty
2 Posts
0
January 19th, 2015 02:00
Thank you Tadhg, for your answer.
You have a goog understand of my issue.
Unfortunately, no solution or workaround, to force Device to be available on site 2.
Regards,
Paul
TadhgConcannon
1 Rookie
1 Rookie
•
22 Posts
0
January 19th, 2015 03:00
Hi Paul
There is no solution or workaround to force Device to be available on Site 2. From what I understand this is because VPLEX Metro believes that Site 1 is available
The only way I think would be to break the WAN comm link first then break the SAN switches
VPLEX preference rules will first first fail to the correct site then break the SAN switches. Must admit I have not tried this though
Be interesting to see what happens
Regards,
Tadhg