PVLINK is active/active link ?!!!?

[Sebastien Masson]

Hi gang

I need your help regarding the LVM pvlink. First of all, my understanding of PVLINK is a active/passive link and not an active/active link like Veristas DMP, this is right or not ?! An email from my HP technical rep. said HP made much improvement in LVM to let pvlink see both paths as active/active link. ?!!!?!

Now all my HP (4 x N4000) are connected together in a SAN with Brocade switches, and all use the same array, a EMC CLARiiON FC-4700. Like any other CLARiiON (Nike, Model 30, etc..) that kind of array is a active/passif array because a LUN can only be access by one SP at a time. If the host tries to reach the LUN by the other link, the other SP connected to it will request it and this is called a trespass.

So trespass append each time I do an ioscan, vgscan, etc... and normally the array support that. But since a upgrade my four serveurs to 11i, my CLARiiON box crash three time (both storage processor panic at the same time, with lost of write cache so data corruption). After analyse, EMC said my four box do too many ioscan/vgscan and what else ?! So too much trespass. Now they are currently try to release a new firmware to handle that better but during this time I want to know if HP-UX 11i does automatic ioscan/vgscan.

EMS is installed and I need it to monitor the hardware and my four node is also part of a cluster (MC ServiceGuard). In the event log of my CLARiiON box I saw many many many over 40 event regarding trespass of LUN and that append every 3 hour constantly (12h00 3h00 6h00 9h00 12h00....). I disable the CHK_IOSCAN_TIME (put value to 0) in diagmond.cfg file.

Fell free to comment any part of that... your help is very appreciate.

[melvyn burnard]

In LVM, a pvlink is an active/passive link.
you would need to ensure that you spread your LUN's across the links to get the best throughput.

[Tom Geudens]

Hi,
I don't know if this is going to help you with your current problem, but you should know that HP has not "certified" the ClarIIon for EMS.
It is recommended to disable monitoring on the ClarIIon devices.
root/sv00247#/etc/opt/resmon/lbin/monconfig
K(ill)
root/sv00247#vi /var/stm/data/tools/monitor/disabled_instances
/storage/events/disks/default/0_3_0_0.8.0.0.0.2.0
/storage/events/disks/default/0_3_0_0.8.0.0.0.1.7
/storage/events/disks/default/0_3_0_0.8.0.0.0.1.6
/storage/events/disks/default/0_3_0_0.8.0.0.0.1.5
/storage/events/disks/default/0_3_0_0.8.0.0.0.1.4
/storage/events/disks/default/0_3_0_0.8.0.0.0.1.3
/storage/events/disks/default/0_3_0_0.8.0.0.0.1.2
/storage/events/disks/default/0_3_0_0.8.0.0.0.1.1
/storage/events/disks/default/0_3_0_0.8.0.0.0.1.0
/storage/events/disks/default/0_3_0_0.8.0.0.0.0.7
/storage/events/disks/default/0_3_0_0.8.0.0.0.0.6
/storage/events/disks/default/0_3_0_0.8.0.0.0.0.5
/storage/events/disks/default/0_3_0_0.8.0.0.0.0.4
/storage/events/disks/default/0_3_0_0.8.0.0.0.0.3
/storage/events/disks/default/0_3_0_0.8.0.0.0.0.2
/storage/events/disks/default/0_3_0_0.8.0.0.0.0.1
/storage/events/disks/default/0_3_0_0.8.0.0.0.0.0
root/sv00247#/etc/opt/resmon/lbin/monconfig
E(nable)

Replace the devices with those used on your system ... EMS will no longer (io)scan them !!!

Hope this helps,
Tom Geudens

[MARTINACHE]

Hi,

We are using thep product "powerpath" from emc in order to use more than one link at the same time.

Ask EMC for the compatibility of the product with your config.

Regards,

Patrice.

[Carlos Fernandez Riera]

Now all my HP (4 x N4000) are connected together in a SAN with Brocade switches, and all use the same array, a EMC CLARiiON FC-4700. Like any other CLARiiON (Nike, Model 30, etc..) that kind of array is a active/passif array because a LUN can only be access by one SP at a time. If the host tries to reach the LUN by the other link, the other SP connected to it will request it and this is called a trespass.

>> On my understand, ioscan can request >>devices to identify, but unless SP must
>>access their disks/cache for data, i dont
>>understand it is a trespass.


>> Every way, tresspass can be caused on both
>>sides, i mean if an SP is not responding LVM
>>will retry by the other way, in facts it
>>produces tresspass.


So trespass append each time I do an ioscan, vgscan, etc... and normally the array support that. But since a upgrade my four serveurs to 11i, my CLARiiON box crash three time (both storage processor panic at the same time, with lost of write cache so data corruption). After analyse, EMC said my four box do too many ioscan/vgscan and what else ?! So too much trespass. Now they are currently try to release a new firmware to handle that better but during this time I want to know if HP-UX 11i does automatic ioscan/vgscan.


>> Maybe increasing timeout for each PV ( see
>> pvchange) will help, but, in the other hand
>>it could be the cause of stoping access to
>>disks from time to time.

>> On a first view it seem to be caused by the upgrade of systems but, as in any other shared devices/chanels, a fail controler/ bad link will cause lots of retrys for all the others.

[Martin Burnett_2]

Hello Sebastian,

EMS versions after Dec01:
B.10.20.24.xx
B.11.00.19.xx
B.11.11.05.xx

No longer monitor disks that do not contain HP firmware (this would include your Clariion disks, DGC), or High Availability disk arrays that already have sufficient monitoring and correction built in (EMC, XP512, etc).

Thanks for participating in the forums,

Martin
Chaos reigns within. Reflect, repent, and reboot. Order shall return.

[Martha Mueller]

We, too, have had problems with errors being logged in syslog.log. In our case, we have tresspass errors during the battery test. The big problem we had occurred when we had a disk fail. The version of the firmware on the fc4700 when it was installed last summer was unable to handle two hot spares. The disk failure caused the raid group holding the failed disk to have such problems that we had to end up rebuilding the raid group. The data wasn't corrupted, though, and we were able to save it off to tape before rebuilding the raid group. We were down for 9 hours. EMC loaded a new version of the firmware to prevent that problem from happening again, but that introduced the errors every time the battery test runs, or running support tools - stm. I have mirrored the data of one of the servers back to internal disks, just to be safe (now that is a real performance hit). It's good to hear about your warnings of data corruption before I unmirror the internal disks. I think I'll wait for the new version of the firmware before breaking the mirror.

[Sebastien Masson]

Just to inform all those that have a FC4700 that run with HP-UX that EMC will release a new Firmware in the next day that is to suppose to fix up a lot the problems linked with HP-UX (trespass, etc).

[Martha Mueller]

Thank you for that update. We have called to schedule the install of the new patch at 07:00 tomorrow. Wish us luck.