This post is more than 5 years old
4 Posts
0
4017
November 4th, 2011 21:00
ESX host tries to read VCMDB device
I have an ESX host that keeps trying to read from VCMDB device.
Vcenter performance metrics are horrible obviously (4000 ms read latency) since it will never actually talk to it.
I have to explain why ESX host is behaving like this just now after months of not reporting any problems.
Is this normal ESX OS behavior?
ESX host is not admin server (No Symcli etc)
This device is controlled by NMP, Powerpath ignores it.
I know I can unmap it from FAs but I want to understand host behavior.
Vsphere 4.1
DMX 4 and Vmax
VCMDB Device 42 is 5.6mb
Thanks !
No Events found!
lion80
4 Posts
0
November 9th, 2011 19:00
I was working with our Vmware team and I suggested rebooting the ESX host, this fixed the issue.
Server now ignores the vcmdb device. No more performance alerts.
Thanks Everybody !
lion80
4 Posts
0
November 4th, 2011 22:00
I know what the VCMDB device is and what it does.
I know it does not impact the customer.
What I don't know is why the host keeps trying all day to talk to it directly.
Management wants an explanation
Jingyi1
2 Intern
•
199 Posts
0
November 6th, 2011 22:00
Well, as I know, VMAX has no VCMDB device. It's using new auto provisioning feature. It's EMC's best practice to only map vcm device to admin workstation. I'm just wondering if you have assigned enough gatekeeper devices for ESX? If not, does ESX use this VCM in the manner of the gatekeeper? This is just my wild guess. The best way to sort it out is to open a ticket as you have severe performance impact.
mattcowger1
61 Posts
0
November 8th, 2011 07:00
While VMAX doesn't have a VCMDB, it does have a similar device called ACLX.
Further, if this host is not an admin server, it should not have ANY gatekeepers or ACLX devices visible to it. Our general recommendation would be to only have ACLX/GKs visible to hosts that should be doing admin functions.
The host keeps trying to talk to it because it keeps looking for a VMFS on the volume - this is pretty normal behavior for ESX, to be honest. Why you've never seen it do this before, I can't answer, but I've seen it happen in my environments before. The way to fix it is to unmap it from the FAs.
dynamox
9 Legend
•
20.4K Posts
0
November 8th, 2011 08:00
ohh trust me, i see my bin file in symmwin before it gets installed. CE tends to put VCM on every single FA so no matter which FA you zone to, you can beging your initial provisioning (masking gatekeeprs and such)
mattcowger1
61 Posts
0
November 8th, 2011 08:00
Your BIn file may not have been configured to expose the VCDM/ACLX on all FA ports - its entirely dependent on how it was configured when installed.
As far as removing a VCMDB/ACLX from production FAs - yes, please be careful and aware of what you are doing. ESX wont care, and if you are following best practices of having only ESX hosts on a given FA port, then it wont be a problem, but make sure you fully understand the impact of any change before you do it.
mattcowger1
61 Posts
0
November 8th, 2011 08:00
I was in your camp too, as a customer. "Dont put anything on my systems that I haven't seen".
Many CEs do put VCM/ACLX/GK on every port just to be safe (hence what the OP is seeing), but some dont. As we move towards VMAX + SMC use, I'm hearing that more and more CE's only put VCM/GKs on 1 set of FA ports, not all (which, in my opinion, is the right move).
dynamox
9 Legend
•
20.4K Posts
0
November 8th, 2011 08:00
we've been running ESX on DMX3 and VMAX and i have never seen these nor have i removed VCMDB/ACLX devices. I would be very careful with removing VCMDB/ACLX on FAs that have been in production. All systems are seeing that device so if you unmap it from FA it will also disappear from all systems connected to that FA ..some may not handle it well as native device names may shift.
dynamox
9 Legend
•
20.4K Posts
0
November 8th, 2011 09:00
in my shop i am actually ok with having VCM/ACLX on each FA, we don't boot from SAN so VCM does not get in the way. Also i have noticed with the latest SE if you run "symcfg discover" it will fail unless you have either VCM or defined gatekeepers. It will not use regular devices for some reason. I am not going to present gatekeepers to every single system that needs to run symcfg discover from time to time ..so having VCM is not a bad thing for me.
SKT2
2 Intern
•
1.3K Posts
0
November 12th, 2011 16:00
also ESX host required V bit disabled from what i remember, so please check that too