Have to remount the SAN after Fiber Cable Failover via MPXIO

[Zahid_Haseeb]

Environment

Two nodes connected with IBM DS8000 SAN

nodes OS = Solaris SPARC 5.9 (Solaris 9)

SAN connected/mounted with nodes via dual Fiber Optics cable

FC cable failover time around 1 minute and 52 seconds

The SAN drive is mounted on /test folder.

expected version of sfha will be install = 5.1 sp1 

Problem

node 1 is tried to copy a file to a SAN Drive mounted on node2. While copying (just for checking) we unplugged active Fiber cable and the copying of file gets stuck. After 1 minute and 52 seconds the failover of fiber cables occurs and the copying which got stuck resumes and again continue copying and finally copied the whole file but with an error. See the error below for reference.

root@PRI-NODE # scp VRTS_SF_HA_Solutions_5.1_SP1_Solaris_SPARC.tar.gz sfha-sol_sparc-5.1SP1RP1-patches.tar.gz SEC-NODE:/test

VRTS_SF_HA_Solutions 100% |*****************************| 639 MB 01:46 

sfha-sol_sparc-5.1SP 100% |*****************************| 121 MB 00:25 

root@PRI-NODE # scp VRTS_SF_HA_Solutions_5.1_SP1_Solaris_SPARC.tar.gz sfha-sol_sparc-5.1SP1RP1-patches.tar.gz SEC-NODE:/test

VRTS_SF_HA_Solutions 100% |*****************************| 639 MB 01:39 

scp: /test/VRTS_SF_HA_Solutions_5.1_SP1_Solaris_SPARC.tar.gz: truncate: I/O error

scp: /test/sfha-sol_sparc-5.1SP1RP1-patches.tar.gz: I/O error

Second after the file copied with an error I noticed that the mounted folder(which is /test in our case) is also not accessable. See the below for reference.

under the /test folder

root@SEC-NODE # ls -l

.: I/O error

Third I also tried again to copy the same file but at this time I was not able to copy the file and after 10minutes I had to unmount the /test folder and mount the SAN again to the /test folder now I am able to copy the files to the /test folder and also able to run the command of ls -l /test.

Question

1.) Why I face the I/O error at the completion of file copied even the  failover of FC occured and file resume the copying after the FC cable unplugged. (just comments required although the veritas product is not installed yet)

2.) why I have to remount the SAN on folder /test. Does this a usual behaviour ? (just comments required although the veritas product is not installed yet)

3.) The main question is that if we install the Storage Foundation, does the vxDMP resolve the above issues ? OR the failover of FC without Veritas installation should be working fine without any I/O Errors?

[g_lee]

What device was used to mount /test? (ie: /dev/dsk/cXtXdXsX ?)

If an MPxIO device was being used, what load balance policy was in use? (eg: round-robin, none)

Please provide the following output

# df -k /test

# cat /kernel/drv/scsi_vhci.conf  ## or fp.conf, depending if mpxio/STMS has been configured globally/per-port

# luxadm display <device>

Regarding whether DMP would prevent the above issue; if the array/device was configured in active/active mode (which is supported for IBM DS8000 per the HCL), then DMP should still be able to write to the disk using the remaining path(s), so would expect it to continue working - you would still get i/o errors for the underlying path but as long as the other paths were up it should still be able to write.

[Marianne]

Please have a look at the following Articles / White Papers / Video to see how exactly DMP handles error detection, path failover, Pro-active Failure Handling, Path Analysis, etc. (Seems you've already seen the video...)

http://www.symantec.com/connect/articles/veritas-storage-foundation-50-dynamic-multi-pathing-optimizing-availability-and-performance

http://www.symantec.com/connect/articles/dynamic-multi-pathing-competetive-comparisons-technical-white-paper

http://www.symantec.com/connect/videos/reduce-downtime-improve-performance-and-increase-scalability-dynamic-multi-pathing-veritas-st

[TonyGriffiths]

Hi,

A few queries if I may.

1) In your test, is Storage Foundation stack being used or not ?

2) Are there any errors in syslog from the Filesystem, MPxIO or SD drivers ?

In theory, a correctly configured and working multipath solution should be able to tolerate loss of a redundant link without any errors at the FS level. The intent of our DMP component is provide such a solution.

cheers

tony

[Zahid_Haseeb · Accepted solution]

Thanks all for their kind replies. Sorry for replying late.

I just did one thing. I installed Storage Foundation and HA. Then I simply created a volume on SAN and mount it on the Secondary Node. I then start copying again the tar file from Primary Node to Secondary Node via SCP command and it got successful.