>
> Warren Wang
>
>
>
> On 6/24/16, 10:23 AM, "Wade Holler" <wade.hol...@gmail.com> wrote:
>
>>On the vm.vfs_cace_pressure = 1 : We had this initially and I still
>>think it is the best choice for most configs. However with our large
>>memory
ph.com
> Cc: Somnath Roy; Warren Wang - ISD; Wade Holler; Blair Bethwaite; Ceph
> Development
> Subject: Re: [ceph-users] Dramatic performance drop at certain number of
> objects in pool
>
>
> Hello,
>
> On Thu, 23 Jun 2016 22:24:59 + Somnath Roy wrote:
>
>> Or
No. Our application writes very small objects.
On Wed, Jun 22, 2016 at 10:01 PM, Blair Bethwaite
<blair.bethwa...@gmail.com> wrote:
> On 23 June 2016 at 11:41, Wade Holler <wade.hol...@gmail.com> wrote:
>> Workload is native librados with python. ALL 4k objects.
>
&g
rkload? I wonder whether for certain workloads (e.g.
> RBD) it's better to increase default object size somewhat before
> pushing the split/merge up a lot...
>
> Cheers,
>
> On 23 June 2016 at 11:26, Wade Holler <wade.hol...@gmail.com> wrote:
>> Based on everyones suggest
ectory is under the
>> calculated threshold and a write occurs (maybe a read, I forget).
>>
> If it's a read a plain scrub might do the trick.
>
> Christian
>> Warren
>>
>>
>> From: ceph-users
>> <ceph-users-boun...@lists.ceph.com<mailto:ceph-users-bou
Thanks everyone for your replies. I sincerely appreciate it. We are
testing with different pg_num and filestore_split_multiple settings. Early
indications are well not great. Regardless it is nice to understand
the symptoms better so we try to design around it.
Best Regards,
Wade
On Mon,
e look like (including slab)?
>
> Cheers,
>
> On 16 June 2016 at 22:14, Wade Holler <wade.hol...@gmail.com> wrote:
>> Hi All,
>>
>> I have a repeatable condition when the object count in a pool gets to
>> 320-330 million the object write time dramatically and
short periods of high read IOPS where (almost) no
> writes occur? What does your memory usage look like (including slab)?
>
> Cheers,
>
> On 16 June 2016 at 22:14, Wade Holler <wade.hol...@gmail.com> wrote:
> > Hi All,
> >
> > I have a repeatable conditi
Hi All,
I have a repeatable condition when the object count in a pool gets to
320-330 million the object write time dramatically and almost instantly
increases as much as 10X, exhibited by fs_apply_latency going from 10ms to
100s of ms.
Can someone point me in a direction / have an explanation ?
We (my customer ) are trying to test at Jewell now but I can say that the
above behavior was also observed by my customer at Infernalis. After 300
million or so objects in a single bucket the cluster basically fell down as
described above. Few hundred osds in this cluster. We are concerned that
___
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
Hi Ben,
What OS+Version ?
Best Regards,
Wade
On Tue, May 3, 2016 at 2:44 PM Ben Hines wrote:
> My crush map keeps putting some OSDs on the wrong node. Restarting them
> fixes it temporarily, but they eventually hop back to the other node that
> they aren't really on.
>
> Is
Hi All,
I searched google and what not but haven't found this yet.
Does anyone know how to do PG -> applicable RadosGW Object mapping?
Best Regards,
Wade
___
ceph-users mailing list
ceph-users@lists.ceph.com
Hi there,
What is the best way to "look at the rgw admin socket " to see what
operations are taking a long time ?
Best Regards
Wade
On Mon, Feb 8, 2016 at 12:16 PM Gregory Farnum wrote:
> On Mon, Feb 8, 2016 at 8:49 AM, Kris Jurka wrote:
> >
> > I've been
a.v...@gmail.com> wrote:
> Am 03.02.2016 um 17:24 schrieb Wade Holler:
> > AFAIK when using XFS, parallel write as you described is not enabled.
> Not sure I'm getting this. If I have multiple OSDs on the same NVMe
> (separated by different data-partitions) I have multiple parallel write
Cheers
Wade
On Thu, Feb 4, 2016 at 7:24 AM Sascha Vogt <sascha.v...@gmail.com> wrote:
> Hi,
>
> Am 04.02.2016 um 12:59 schrieb Wade Holler:
> > You referenced parallel writes for journal and data. Which is default
> > for btrfs but but XFS. Now you are mentioning multiple
Hi Sascha,
What is your file system type, XFS or Btrfs ?
Thanks
Wade
On Wed, Feb 3, 2016 at 7:01 AM Sascha Vogt wrote:
> Hi all,
>
> we recently tried adding a cache tier to our ceph cluster. We had 5
> spinning disks per hosts with a single journal NVMe disk, hosting
but if your KVM instances are really that short
lived, could you get away with size=2 on the cache pool from and
availability perspective ?
On Wed, Feb 3, 2016 at 7:44 AM Sascha Vogt <sascha.v...@gmail.com> wrote:
> Hi Wade,
>
> Am 03.02.2016 um 13:26 schrieb Wade Holler:
> > Wha
Could you share the fio command and your read_ahead_kb setting for the OSD
devices ? "performance is better" is a little too general. I understand
that we usually mean higher IOPS or higher aggregate throughput when we say
performance is better. However, application random read performance
327.3.1 as I recall. Thank for the additional information / confirmation.
On Mon, Feb 1, 2016 at 6:05 PM Simon Ironside <sirons...@caffetine.org>
wrote:
> On 01/02/16 17:47, Wade Holler wrote:
> > I can at least say that I've seen this. (a lot)
> >
> > Running Infern
I can at least say that I've seen this. (a lot)
Running Infernalis with Btrfs on Cent 7.2. I haven't seen any other issue
in the cluster that I would say is related.
Take it for what you will.
Best Regards,
Wade
On Mon, Feb 1, 2016 at 12:39 PM Simon Ironside
wrote:
Great commentary.
While it is fundamentally true that higher clock speed equals lower
latency, I'm my practical experience we are more often interested in
latency at the concurrency profile of the applications.
So in this regard I favor more cores when I have to choose, such that we
can support
> /dev/sdb[2,3,4] and I was able to start all osd just running systemctl
> start ceph.target
>
> Cheers
> Goncalo
>
>
> --
> *From:* Wade Holler [wade.hol...@gmail.com]
> *Sent:* 08 January 2016 01:15
> *To:* Goncalo Borges; Loic Dacha
appropriate ?
Thank you ahead of time for your help!
Best Regards,
Wade
On Mon, Jan 11, 2016 at 5:43 PM John Spray <jsp...@redhat.com> wrote:
> On Mon, Jan 11, 2016 at 10:32 PM, Wade Holler <wade.hol...@gmail.com>
> wrote:
> > Does anyone else have any suggestions here? I a
:12 AM Wade Holler <wade.hol...@gmail.com> wrote:
> It is not set in the conf file. So why do I still have this behavior ?
>
> On Fri, Jan 8, 2016 at 11:08 AM hnuzhoulin <hnuzhoul...@gmail.com> wrote:
>
>> Yeah,this setting can not see in asok config.
>> You just
t in the crushmap, even if the crushmap does not reflect
> reality.
>
> Regards,
>
> Mart
>
>
>
>
>
> On 01/08/2016 02:16 AM, Wade Holler wrote:
>
> Sure. Apologies for all the text: We have 12 Nodes for OSDs, 15 OSDs per
> node, but I will only include a
My experience is performance degrades dramatically when dirty objects are
flushed.
Best Regards,
Wade
On Fri, Jan 8, 2016 at 11:08 AM hnuzhoulin wrote:
> Hi,guyes
> Recentlly,I am testing cache-tier using writeback mode.but I found a
> strange things.
> the performance
f these restart is necessary)
>
> what I use this config is when I changed crushmap manually,and I do not
> want the service init script to rebuild crushmap as default way.
>
> maybe this is not siut for your problem.just have a try.
>
> 在 Fri, 08 Jan 2016 21:51:32 +0800,Wa
}
host cpn4 {
Thank you for your review !
Wade
On Thu, Jan 7, 2016 at 6:03 PM Shinobu Kinjo <ski...@redhat.com> wrote:
> Can you share the output with us?
>
> Rgds,
> Shinobu
>
> - Original Message -----
> From: "Wade Holler" <
I commented out partprobe and everything seems to work just fine.
*If someone has experience with why this is very bad please advise.
Make sure you know about http://tracker.ceph.com/issues/13833 also.
*ps we are running btrfs in the test jig and had to add the "-f" to the
btrfs_args for
Sometimes my ceph osd tree output is wrong. Ie. Wrong osds on the wrong
hosts ?
Anyone else have this issue?
I have seen this at Infernalis and Jewell.
Thanks
Wade
___
ceph-users mailing list
ceph-users@lists.ceph.com
All,
I am testing an all SSD and NVMe (journal) config for a customers first
endeavor investigating Ceph for performance oriented workloads.
Can someone recommend a good performance and reliable ( under high load )
combination?
Terrible high level question I know but we have had a number of
Have a look at the iostsat -x 1 1000 output to see what the drives are doing
On Wed, Dec 23, 2015 at 4:35 PM Florian Rommel <
florian.rom...@datalounges.com> wrote:
> Ah, totally forgot the additional details :)
>
> OS is SUSE Enterprise Linux 12.0 with all patches,
> Ceph version 0.94.3
> 4
I had major host stability problems under load with -327 . Repeatable test
cases under high load with XFS or BTRFS would result in hung kernel tasks
and of course the sympathetic behavior you mention.
requests are blocked mean that the op tracker in ceph hasn't received a
timely response from
ge, maybe there is a sysctl option to tweak on OSDs ? this
> will be nasty if it goes into our production!
>
>
>
> --
>
> Dan
>
>
>
> *From:* Wade Holler [mailto:wade.hol...@gmail.com]
> *Sent:* Tuesday, December 22, 2015 4:36 PM
> *To:* Dan Nica <dan.n...@staff.bluemat
Hi Dan ,
When we say "mount" we are usually referring to a file system. Mounting a
non shared filesystem on multiple hosts concurrently will certainly break
things since each non shared filesystem host thinks it has exclusive
access. Of course this is not true if a shared / clustered filesystem
I'm interested in this too. Should start testing next week at 1B+ objects
and I sure would like a recommendation of what config to start with.
We learned the hard way that not sharding is very bad at scales like this.
On Wed, Dec 16, 2015 at 2:06 PM Florian Haas wrote:
> Hi
Keep it simple is my approach. #1
If needed Add rudimentary HA with pacemaker.
http://linux-ha.org/wiki/Samba
Cheers
Wade
On Tue, Dec 15, 2015 at 5:45 AM Alex Leake wrote:
> Good Morning,
>
>
> I have a production Ceph cluster at the University I work at, which runs
>
38 matches
Mail list logo