Unsolved
This post is more than 5 years old
7 Posts
0
2797
Attach FreeBSD host to VNX5300 SAN
Hello,
FreeBSD is not officially supported by EMC as an host OS, so I'm a bit on my own with this problem. I'm trying to use FreeBSD (9.3) on HP blade servers and use our VNX5300 as a SAN storage with these hosts.
Everything looks like its OK, but in fact performances are horrendous as FC paths won't work as expected. When I try to read or write on the multipath device created on FreeBSD, the first path (on first scsi bus) is always faulty: I got seek times like 250 ms, and read transfer rate like 2-300 kB/s.
Eventually this path will fail, and a path on the second scsi bus will be picked. Then seek times are around 3 ms and transfer rates around 100 MB/s (ultra basic bench made with diskinfo).
Logs are filled with errors:
isp0: Chan 0 Abort Cmd for N-Port 0x0008 @ Port 0x090a00
(da2:isp0:0:2:0): Command Aborted
(da2:isp0:0:2:0): WRITE(10). CDB: 2a 00 00 50 20 11 00 00 07 00
(da2:isp0:0:2:0): CAM status: CCB request aborted by the host
(da2:isp0:0:2:0): Retrying command
and the multipath device ends up in a degraded state:
# gmultipath status
Name Status Components
multipath/SPLUNK_1 DEGRADED da3 (PASSIVE)
da7 (FAIL)
da2 (ACTIVE)
da6 (FAIL)
Is there any working settings in Unisphere that I can use to properly interact with a FreeBSD host? (I can upgrade to FreeBSD 10.x if necessary).
Is there any tuning I can do on FreeBSD's side too?
thanks,
Patrick
dynamox
2 Intern
2 Intern
•
20.4K Posts
0
December 10th, 2014 04:00
have you tried a different failover mode on VNX, maybe instead of 4 try 1 ?
patpro
7 Posts
0
December 10th, 2014 05:00
I did. FreeBSD won't even boot entirely with that setting. Console would display something like this:
run_interrupt_driven_hooks still waiting after 300 seconds forxpt_config
May be it's unrelated, but it happened only with this setting.
I've destroyed my multipath device on the FreeBSD host, unregistered all its initiators in Unisphere and recreated them with failover 1. Then I've recreated the multipath device. Now the server boots properly. The device uses only 2 providers:
# gmultipath status
Name Status Components
multipath/SPLUNK_1 OPTIMAL da2 (ACTIVE)
da6 (PASSIVE)
But my performance problem is still there:
Seek times:
Full stroke: 250 iter in 60.601416 sec = 242.406 msec
Half stroke: 250 iter in 61.690406 sec = 246.762 msec
and errors are present in logs:
isp0: Chan 0 Abort Cmd for N-Port 0x0008 @ Port 0x090a00
(da2:isp0:0:2:0): Command Aborted
(da2:isp0:0:2:0): READ(10). CDB: 28 00 0c 7f fe a1 00 00 01 00
(da2:isp0:0:2:0): CAM status: CCB request aborted by the host
(da2:isp0:0:2:0): Retrying command
isp0: Chan 0 Abort Cmd for N-Port 0x0008 @ Port 0x090a00
(da2:isp0:0:2:0): Command Aborted
(da2:isp0:0:2:0): READ(6). CDB: 08 00 1c d9 f8 00
(da2:isp0:0:2:0): CAM status: CCB request aborted by the host
(da2:isp0:0:2:0): Retrying command
dynamox
2 Intern
2 Intern
•
20.4K Posts
0
December 10th, 2014 08:00
you are not booting from VNX, are you ?
patpro
7 Posts
0
December 10th, 2014 14:00
No, that's not reliable enough, I won't even try. The server boots from an SD card inside, seen as an USB device.
Alex_Ye
109 Posts
0
December 27th, 2014 20:00
This is an abort. I'd not think it's a mis-configuration.
Since you are seeing issues on the "first scsi bus", could it be a bad cable, SFP, etc.?
patpro
7 Posts
0
January 5th, 2015 06:00
I think the hardware is OK. The enclosure holds 16 blade servers, and many of them are still production (VMware or dedicated Linux). I have no indication that my 2 test blades are defective (recently decommissioned from production). Also, the same blade works perfectly when I boot a RHEL or a Centos 6.
rawstorage
419 Posts
0
January 5th, 2015 07:00
Since what your are trying to do here is totally unsupported I would question the why?, In the event you have more problems down the line you will not be able to get any support on this configuration. Why not try with a supported version of redhat even if you don't have full OS vendor support you will at least know that the config should work and there is a lot of supported documentation out there to help you get going.
Just because you could doesn't mean you should.
patpro
7 Posts
0
January 5th, 2015 08:00
Because I really do prefer working on FreeBSD, for a lot of reasons. In the mean time I've installed a couple of CentOS to host my Splunk deployment, but as soon as I can switch back to FreeBSD, I will.
patpro
7 Posts
0
January 6th, 2015 05:00
Could you clarify what are "RPQ" and "eLab"? Thanks
rawstorage
419 Posts
0
January 6th, 2015 05:00
Patpro,
unfortunately this won't get you any closer to a supported environment perhaps you could submit an RPQ with your local team for support and guidance from eLab might get you closer
rawstorage
419 Posts
0
January 6th, 2015 06:00
RPQ is a special process to give support for environments not explicitly supported in the EMC Support Matrix, your account team will be able to handle this for you.
eLAB is essentially the EMC support Matrix.
patpro
7 Posts
0
January 6th, 2015 06:00
OK, thank you. I'll take a look into that RPQ process as soon as possible.