On 11-8-2016 13:44, Wido den Hollander wrote:
> Could be, but the log line which is interesting:
>
> 2016-08-11 12:06:16.473351 b69c480 10 osd.0 178 handle_osd_ping osd.2
> 127.0.0.1:6811/13222 says i am down in 178
So I've augmented the code a bit:
case MOSDPing::YOU_DIED:
dout(10) << "handle_osd_ping " << m->get_source_inst()
<< " says i am down in " << m->map_epoch << dendl;
dout(10) << "handle_osd_ping subscribing to: " << curmap->get_epoch()+1
<< dendl;
osdmap_subscribe(curmap->get_epoch()+1, false);
break;
And we go into odsmap_subscribe, which I augmented too:
void OSD::osdmap_subscribe(version_t epoch, bool force_request)
{
OSDMapRef osdmap = service.get_osdmap();
if (osdmap->get_epoch() >= epoch) {
dout(10) << __func__ << " subscribing to: " << epoch
<< " but osdmap already has: " << osdmap->get_epoch()
<< dendl;
return;
}
if (monc->sub_want_increment("osdmap", epoch, CEPH_SUBSCRIBE_ONETIME)
||force_request) {
dout(10) << __func__ << " renewing subscription" << dendl;
monc->renew_subs();
}
}
If I run with force_request = false, then this call does absolutely
nothing... Neither if's trigger
If I set force_request = true (expecting it to fetch the whole map),
that does not do much either.
It reports:
monclient: renew_subs - empty
So either this path is never/rarely used, or is not the solution to the
problem.
--WjW
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
