EVA 4400 (xcs 09501000)

[Dja_1]

We have a pair of EVA4400s not quite in production. On "EVA1" we have been migrating some data to the array and tonight had increased the size on a Vdisk from 1500 to 2000GB. At that point (after clicking Ok) the array was no longer accesible. All connected hosts lost sight of disks and the EVA is not now accessible from the SMS (CV9) nor the integrated MM. The WOCP was available although I could not do anything useful with it. The WOCP does show one controller as good but that seems not to be the case. Attempts to restart a controller or the system from WOCP gets me "Command execution failed. Error: Controller API Not Ready, Try Again Later". I have shut down everthing and brought back online but get to the same hung state. xcs = 09501000. Any ideas???

[Dileep..]

Do log a case with HP. this may need extensive log analysis.

[Dja_1]

Oddly.... after approx 8 hours it has come back to life without any intervention. I will have HP on site asap next week.

[erkle]

I am not happy at all with the EVA4400. I have implemented about 5 EVA4400 for our customers over past 12 months and have serious reservations about the stability of the new HSV300 controllers. Numerous issues with controllers locking up, high rate of failing 1TB drives, loosing access to Command View EVA, and issues with grouping 1TB drives with double protection.
I think HP engineering need to seriously road test new kit (especially EVA) before releasing to market. The amount of firmware updates for EVA4400 in past 9 months proves that HP are still fine tuning the HSV300 controllers.
I am also starintg to doubt HP on their tech support for EVA.

[Dja_1]

Interesting comment. I must say I am less confident about the migration of production systems to this array. Its mirror 4400 at our HO has the previous firmware and has been stable. Not sure if there is any significance there. I should add to my last post that although the problem 4400 is reporting itself as "good" i cannot access any of the existing vdisks from owning hosts and boot from san hosts will not boot at all. thankfully these are test hosts or Citrix Xen where there is plenty of redundancy elswhere in the network. Last night I tried to present a new vdisk from the problem array to a new host (w2k3) and it hung the OS. An interesting week ahead.

[Dja_1]

I have just tried a graceful shutdown of the array from the WOCP. Doesn't power off disk enclosures.... I have noted this was an issue with older firware releases but seems it persists. Can someone advise the next best approach to graceful power down of the enclosures? Thanks.

[SAKET_5]

Dja,

We have had a controller lock up issue with one of our EVA4400s - a "Vdisk" stalled for over 60 minutes and then master controller status conflicts occured - XCS firmware then instructed one controller to shutdown. This incident caused a few Linux hosts to crash and an issue with a couple of unmounted filesystems on a Windows/PolyServe cluster.

HP support initially advised us that XCS 09501000 got released just last month which addresses a number of these bugs. "A more recent advice from HP to us was to hold off on xcs 09501000 upgrade as HP has put this version on freeze now." Did you upgrade to xcs 09501000 just recently and was it based on HP's advice or did you proactively perform this upgrade?

Regards,
Saket

[SAKET_5]

Dja,

To the best of my understanding, the graceful EVA shutdown options can be found either via CV-EVA, CV-EVA field service page has a fast reboot option (not shutdown), integrated MM or the OCP. You seem to have exhausted all these options. My recommendation would be to let HP guys come and have a look at the system in its current state.

[Dja_1]

It was a proactive upgrade in-house.... wanting to be at latest release (given the original strong public rec from HP to update) before going live. We will be looking for a regression path if the issues are not resolved soon. p.s. without powering of the disk enclosures the graceful restart of the controllers resolved the remaing issues.

[erkle]

I had the same issue with lost access to CV when unpresenting a vdisk from a host. For approx 60 mins no access to CV but no loss of access to other Luns and hosts were running fine (ESX 3.5). After 60mins controller 2 rebooted and gave errors in log. Also showed "Soft Diagnostic Fasilure" on controller 2 when it came back. When I looked closer, controller2 had a failed cache battery. Hp came onsite reseated the cache battery and it came back ok. When we rebooted controller 2 that evening the soft diagnostic error went away and EVA was green again.