[ovirt-users] High latency on storage domains and sanlock renewal error

[Stefano Bovina]

Hi,
a little update:

The command multipath -ll hung when executed on the host while the problem
occur (nothing logged in /var/log/messages or dmesg).

I tested latency with ioping:
ioping /dev/6a386652-629d-4045-835b-21d2f5c104aa/metadata

Usually it return "time=15.6 ms", sometimes return "time=19 s" (yes,
seconds)

Systems are up to date and I tried both path_checker (emc_clariion and
directio), without results.
(https://access.redhat.com/solutions/139193, it refers to the Rev A31 of
EMC document; last is A42 and suggest emc_clariion).

Any idea or suggestion?

Thanks,

Stefano

2017-05-08 11:56 GMT+02:00 Yaniv Kaul <ykaul at redhat.com>:

>
>
> On Mon, May 8, 2017 at 11:50 AM, Stefano Bovina <bovy89 at gmail.com> wrote:
>
>> Yes,
>> this configuration is the one suggested by EMC for EL7.
>>
>
> https://access.redhat.com/solutions/139193 suggest that for alua, the
> patch checker needs to be different.
>
> Anyway, it is very likely that you have storage issues - they need to be
> resolved first and I believe they have little to do with oVirt at the
> moment.
> Y.
>
>
>>
>> By the way,
>> "The parameters rr_min_io vs. rr_min_io_rq mean the same thing but are
>> used for device-mapper-multipath on differing kernel versions." and
>> rr_min_io_rq default value is 1, rr_min_io default value is 1000, so it
>> should be fine.
>>
>>
>> 2017-05-08 9:39 GMT+02:00 Yaniv Kaul <ykaul at redhat.com>:
>>
>>>
>>> On Sun, May 7, 2017 at 1:27 PM, Stefano Bovina <bovy89 at gmail.com> wrote:
>>>
>>>> Sense data are 0x0/0x0/0x0
>>>
>>>
>>> Interesting - first time I'm seeing 0/0/0. The 1st is usually 0x2 (see
>>> [1]), and then the rest [2], [3] make sense.
>>>
>>> A google search found another user with Clarion with the exact same
>>> error[4], so I'm leaning toward misconfiguration of multipathing/clarion
>>> here.
>>>
>>> Is your multipathing configuration working well for you?
>>> Are you sure it's a EL7 configuration? For example, I believe you should
>>> have rr_min_io_rq and not rr_min_io .
>>> Y.
>>>
>>> [1] http://www.t10.org/lists/2status.htm
>>> [2] http://www.t10.org/lists/2sensekey.htm
>>> [3] http://www.t10.org/lists/asc-num.htm
>>> [4] http://www.linuxquestions.org/questions/centos-111/multi
>>> path-problems-4175544908/
>>>
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/users/attachments/20170512/cbeb3346/attachment.html>