CA warning: 4400 Stan: Excessive Vdisk response time & Copy resources on the inter site link have be

[Magnus Appelgren]

When doing full copy with CA over 2G isl to remote site i get these errors:
---------- Problem Found: CUST: CA informational: 4400 Stan: Copy resources on the inter site link have been reduced.

and

---------- Problem Found: CUST: CA warning: 4400 Stan: Excessive Vdisk response time at the DR Source has been detected.

I have 2 EVA 4400 with fw:09522000.

I'm gonna open a case tomorrow but in the mean time any suggestions would be helpful. Messing a bit with evaperf but not really sure which commands to execute... Vmware vmotion is very slow at the moment and I have a lot more full copies to do...

Regards Magnus

[Víctor Cespón]

EVAperf drg

The message indicates that the EVA where you're replicating your data is not responding fast enough.
This can be due to the EVA itself, the amount of work it's doing and the number of disks in the disk group. Or due to the speed of the replication link.

Before setting up a Continous Access installation, you must collect information and use CA sizing tools.

VmWare + FATA disks + Small disk groups + LUNs in VRAID5 = very bad performance

If you have set up the DR Groups in synchronous mode, then even worst performance.

[Magnus Appelgren]

Both source and destination is 22 x 450gb. vraid 5. Synchrounus mode used. Any other ideas?

Is it very big difference between synchrounus and asynchronous mode?

[Víctor Cespón]

Yes, in synchronous mode a write has to be completed on both EVAs before it's considered done for the host.
In asynchronous mode the writes are done on the source EVA and then it copies the data to the other EVA when it can.

In your case the DR link or the destination EVA are not fast enough to work in synchronous mode.

Synchronous mode has the advantage that the data on both EVAs will be identical in any moment, but you must be able to transfer all the data through the DR link in real time.

[Magnus Appelgren]

What am I supposed to see with evaperf drg? Seems to be all 0 all the time.

Both EVA's are identical in model and fw. The link is 2gb. We have 4 vmware hosts with about 50 servers. Mostly small ones, no clusters ot multi-cpu. 1 exchange and maybe 15 sql. Somethings seems wrong, i think the link should be able to handle this. Dest EVA could have problems that I dont know.

[Herr Gottfried HOELD]

We have similar problems with two EVA4400 and SVSP, so we change the Disk.DiskMaxIOSize to 1024k on the ESX 3.5. Now the write latency of the vdisks decreased.

Verify with evaperf vd -cont

and check the VMware KB
http://kb.vmware.com/selfservice/microsites/search.do?cmd=displayKC&docType=kc&externalId=1003469&sliceId=1&docTypeID=DT_KB_1_1&dialogID=14801522&stateId=0%200%2014807012

[skt_skt]

what is the distance between the sites? is that an OLTP environment?How much GB is transffered on a daily basis? Are you using ay compress mode during the transfer?

I have used CA couple of years back and were syncing the CA pair once in a day and sent to tape. So make sure your link has enough bandwidth to support the rate of change if not you are going to impact your PROD sonner or later .

[Magnus Appelgren]

About 4 kilometers. Whats OLTP? Best way to meassure a days total bw? No compression afaik.

[Uwe Zessin]

OnLine Transaction Processing = small, random I/O

[Magnus Appelgren]

What does "Copy resources on the inter site link have been reduced." really mean? Even after full copy is done it feel very very slow. When will it be back to normal?