All of lore.kernel.org
 help / color / mirror / Atom feed
* cuttlefish punch list
@ 2013-04-29 15:43 Sage Weil
  2013-04-30  4:05 ` Sage Weil
  2013-04-30  7:52 ` kernel 3.9 / new networking options for RADOS Kasper Dieter
  0 siblings, 2 replies; 3+ messages in thread
From: Sage Weil @ 2013-04-29 15:43 UTC (permalink / raw)
  To: ceph-devel

There are two big items I'm concerned about:

#4815, and the other large monitor store problems with syncing and failure 
to get quorum.  We need to confirm
 - why the sync is failing and we're hitting that assert, and
 - why the store.db's are growing so huge

#4853: This is triggered by one of the upgrade tests and looks like it 
could be a regression from the client request replay fix last week.

A few other things:

 - the rgw upgrade tests are failing because the s3tests, readwrite, and 
   swift tasks can't be rung twice in a row.
 - ceph release build needs to include raring.
 - ceph-deploy build needs to include raring too.

Thanks!
sage

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: cuttlefish punch list
  2013-04-29 15:43 cuttlefish punch list Sage Weil
@ 2013-04-30  4:05 ` Sage Weil
  2013-04-30  7:52 ` kernel 3.9 / new networking options for RADOS Kasper Dieter
  1 sibling, 0 replies; 3+ messages in thread
From: Sage Weil @ 2013-04-30  4:05 UTC (permalink / raw)
  To: ceph-devel

On Mon, 29 Apr 2013, Sage Weil wrote:
> There are two big items I'm concerned about:
> 
> #4815, and the other large monitor store problems with syncing and failure 
> to get quorum.  We need to confirm
>  - why the sync is failing and we're hitting that assert, and
>  - why the store.db's are growing so huge

This appears to be a bad interaction between the mon workload and levedb's 
compaction heuristics.  We've worked around it in wip-mon-compact; anyone 
who has seen large mon stores and/or a failure to sync, please try this 
branch!

> #4853: This is triggered by one of the upgrade tests and looks like it 
> could be a regression from the client request replay fix last week.

Fixed.

> A few other things:
> 
>  - the rgw upgrade tests are failing because the s3tests, readwrite, and 
>    swift tasks can't be rung twice in a row.

Running now.

>  - ceph release build needs to include raring.
>  - ceph-deploy build needs to include raring too.

Tomorrow.

Thanks everyone!  We're close!
sage

^ permalink raw reply	[flat|nested] 3+ messages in thread

* kernel 3.9 / new networking options for RADOS
  2013-04-29 15:43 cuttlefish punch list Sage Weil
  2013-04-30  4:05 ` Sage Weil
@ 2013-04-30  7:52 ` Kasper Dieter
  1 sibling, 0 replies; 3+ messages in thread
From: Kasper Dieter @ 2013-04-30  7:52 UTC (permalink / raw)
  To: ceph-devel; +Cc: Kasper Dieter

What are the options for RADOS to make use of the new
networking feature in Kernel 3.9 ?

http://www.h-online.com/open/features/What-s-new-in-Linux-3-9-1845705.html

TCP and UDP sockets now support a SO_REUSEPORT option that allows multiple
sockets to listen on the same port. This enables, for example, multiple
web server processes or threads to open individual sockets to listen
on port 80; any connections that come in on this port will be evenly
distributed across the sockets by the kernel.

This approach aims to improve the way workloads are distributed across the
available processor cores. (...) this extension can help avoid bottlenecks
that appear in certain situations when only one thread accepts new
connections and then distributes them across other threads. (...)


Kind Regards,
-Dieter

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2013-04-30  7:53 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2013-04-29 15:43 cuttlefish punch list Sage Weil
2013-04-30  4:05 ` Sage Weil
2013-04-30  7:52 ` kernel 3.9 / new networking options for RADOS Kasper Dieter

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.