Re: PGLog::rewind_divergent_log use case

[Loic Dachary <loic@dachary.org>]

[-- Attachment #1: Type: text/plain, Size: 2312 bytes --]

Hi Sam,

I took a look at calc_acting and I can't figure out enough to create a bug report. There are checks based on last_update and tail but it is difficult go infer how it relates to the rewind_divergent_log function. If you can connect the dots for me I'd be happy to write down a proper bug report ;-)

Cheers

On 06/11/2013 01:02 AM, Samuel Just wrote:
> I think that case may be possible, and that the assert may be wrong.
> We have probably never seen a problem there due to
> osd_min_pg_log_entries.  It's also possible that there is a subtle bug
> between there and calc_acting().
> -Sam
> 
> On Sat, Jun 8, 2013 at 1:16 AM, Loic Dachary <loic@dachary.org> wrote:
>> [Note: ceph-devel cc'ed]
>>
>> Hi Sam,
>>
>> While trying to create the conditions for the following to happen in PGLog::rewind_divergent_log:
>>
>>     if (p == log.log.begin()) {
>>       // yikes, the whole thing is divergent!
>>       divergent.swap(log.log);
>>       break;
>>     }
>>
>> https://github.com/ceph/ceph/blob/master/src/osd/PGLog.cc#L355
>>
>> I created this ( matching the drawing attached ), and the test case works.
>>
>>     {
>>       pg_log_entry_t e;
>>
>>       info.log_tail = log.tail = eversion_t(1, 1);
>>       newhead = eversion_t(1, 3);
>>       e.version = divergent_version = eversion_t(1, 5);
>>       e.soid.hash = 0x9;
>>       divergent_object = e.soid;
>>       e.op = pg_log_entry_t::DELETE;
>>       e.prior_version = prior_version = eversion_t(0, 2);
>>       log.log.push_back(e);
>>       log.head = e.version;
>>     }
>>
>> https://github.com/dachary/ceph/commit/d592d7f10968c56139cb470a4111ddfbe08a4030#L0R136
>>
>> But it does not look like a situation that is actually going to happen. I feel something is missing and a hint would be much appreciated :-)
>>
>> Cheers
>>
>> --
>> Loïc Dachary, Artisan Logiciel Libre
>> All that is necessary for the triumph of evil is that good people do nothing.
> --
> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

-- 
Loïc Dachary, Artisan Logiciel Libre
All that is necessary for the triumph of evil is that good people do nothing.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 262 bytes --]