Thanks everyone,
update: I tried running on node A:
# vmtouch -ev /storage/
# sync; sync
The problem persisted; one minute needed to 'ls -Ral' the dir (from node B).
After that I ran on node A:
# echo 2 /proc/sys/vm/drop_caches
And everything became suddenly fast on node B. ls, du, tar, all
It is not only ls; even du or tar are extremely slow.
Example with tar (from node B):
# time tar c /storage/test10/installed-tests/pts/pgbench-1.5.1//dev/null
real1m45.291s
user0m0.023s
sys 0m0.143s
While on the node that originally wrote the dir (node A):
# time tar c
On 16/06/2015 12:11, negillen negillen wrote:
This is quite a problem because we have several applications that need
to access a large number of files and when we set them to work on
CephFS latency skyrockets.
The question of what access means here is key. Especially, whether
you need the
Hi!
As I know:
C60X = SAS2 = 3Gbps
LS2308 = 6Gbps
Onboard SATA3 = 6Gbps (usually only 2 ports)
Onboard SATA2 = 3Gbps (4-6 ports)
We use Intel S2600 motherboards and R2224GZ4 platforms in our Hammer
evaluation instance. C60X connected to 4-drive 2.5 bay: 2 small SAS drives for
OS.
2xS3700
Okay…I think the extra layers you have will add some delay, but 1m is high
probably (I never tested Ceph on HDD though).
We can minimize it probably by optimizing the cluster setup.
Please monitor your backend cluster or even the rbd nodes to see if anything is
bottleneck there.
Also, check if
This worked like a champ, thank you kindly!
On Jun 15, 2015, at 2:16 PM, Somnath Roy somnath@sandisk.com wrote:
If you want to suppress the warning, do this in the conf file..
mon_pg_warn_max_per_osd = 0
or
mon_pg_warn_max_per_osd = something big
Thanks Regards
I forgot to ask, is this with the patched version of tcmalloc that
theoretically fixes the TCMALLOC_MAX_TOTAL_THREAD_CACHE_BYTES issue?
Yes, the patched version of tcmalloc, but also the last version from gperftools
git.
(I'm talking about qemu here, not osds).
I have tried to increased
Fixed! At least looks like fixed.
It seems that after migrating every node (both servers and clients) from
kernel 3.10.80-1 to 4.0.4-1 the issue disappeared.
Now I get decent speeds both for reading files and for getting stats from
every node.
Thanks everyone!
On Tue, Jun 16, 2015 at 1:00 PM,
Hi Cephers,
While looking at disk utilization on OSD, I noticed the disk was constantly
busy with large number of small writes, further investigation showed that, as
radosgw uses xattrs to store metadata (e.g. etag, content-type, etc.), which
made the xattrs get from local to extents, which
Guang,
Try to play around with the following conf attributes specially
filestore_max_inline_xattr_size and filestore_max_inline_xattrs
// Use omap for xattrs for attrs over
// filestore_max_inline_xattr_size or
OPTION(filestore_max_inline_xattr_size, OPT_U32, 0) //Override
Hi,
some news about qemu with tcmalloc vs jemmaloc.
I'm testing with multiple disks (with iothreads) in 1 qemu guest.
And if tcmalloc is a little faster than jemmaloc,
I have hit a lot of time the tcmalloc::ThreadCache::ReleaseToCentralCache bug.
increasing
Hi Alexandre,
Excellent find! Have you also informed the QEMU developers of your
discovery?
Mark
On 06/16/2015 11:38 AM, Alexandre DERUMIER wrote:
Hi,
some news about qemu with tcmalloc vs jemmaloc.
I'm testing with multiple disks (with iothreads) in 1 qemu guest.
And if tcmalloc is a
I forgot to ask, is this with the patched version of tcmalloc that
theoretically fixes the TCMALLOC_MAX_TOTAL_THREAD_CACHE_BYTES issue?
Mark
On 06/16/2015 11:46 AM, Mark Nelson wrote:
Hi Alexandre,
Excellent find! Have you also informed the QEMU developers of your
discovery?
Mark
On
On Wed, 17 Jun 2015, Zhou, Yuan wrote:
FWIW, there was some discussion in OpenStack Swift and their performance
tests showed 255 is not the best in recent XFS. They decided to use large
xattr boundary size(65535).
https://gist.github.com/smerritt/5e7e650abaa20599ff34
If I read this
Hi,
I finally fix it with tcmalloc with
TCMALLOC_MAX_TOTAL_THREAD_CACHE_BYTES=268435456 LD_PRELOAD} =
/usr/lib/libtcmalloc_minimal.so.4 qemu
I got almost same result than jemmaloc in this case, maybe a littleb it faster
Here the iops results for 1qemu vm with iothread by disk
After back-porting Sage's patch to Giant, with radosgw, the xattrs can get
inline. I haven't run extensive testing yet, will update once I have some
performance data to share.
Thanks,
Guang
Date: Tue, 16 Jun 2015 15:51:44 -0500
From: mnel...@redhat.com
To: yguan...@outlook.com;
Hi Yuan,
Thanks for sharing the link, it is interesting to read. My understanding of the
test results, is that with a fixed size of xattrs, using smaller stripe size
will incur larger latency for read, which kind of makes sense since there are
more k-v pairs, and with the size, it needs to get
Hi ceph experts,
I did some test on my ceph cluster recently with following steps:
1. at the beginning, all pgs are active+clean;
2. stop a osd. I observed a lot of pgs are degraded.
3. ceph osd out.
4. then I observed ceph is doing recovery process.
the ceph cluster is configured as three
Hello,
On Tue, 16 Jun 2015 07:21:54 + Matteo Dacrema wrote:
Hi,
I've shutoff the node without take any cautions for simulate a real case.
Normal shutdown (as opposed to simulating a crash by pulling cables)
should not result in any delays due to Ceph timeouts.
The
anyone else who can help me with this?
-Original Message-
From: Irek Fasikhov
malm...@gmail.commailto:irek%20fasikhov%20%3cmalm...@gmail.com%3e
To: Makkelie, R (ITCDCC) - KLM
ramon.makke...@klm.commailto:%22Makkelie,%20r%20%28itcdcc%29%20-%20klm%22%20%3cramon.makke...@klm.com%3e
Cc:
Hi,
I've shutoff the node without take any cautions for simulate a real case.
The osd_pool_default_min_size is 2 .
Regards,
Matteo
Da: Christian Balzer ch...@gol.com
Inviato: martedì 16 giugno 2015 01:44
A: ceph-users
Cc: Matteo Dacrema
Oggetto: Re:
Hello,
you're right.
I misunderstood the meaning of the two configuration params: size and min_size.
Now it works correctly.
Thanks,
Matteo
Da: Christian Balzer ch...@gol.com
Inviato: martedì 16 giugno 2015 09:42
A: ceph-users
Cc: Matteo Dacrema
That's expected behaviour. If RADOS can't make your writes safe by
replicating them (because no other OSD is available) then clients will
pause their writes. See the min_size setting on a pool.
John
On 16/06/2015 00:11, Matteo Dacrema wrote:
With 3.16.3 kernel it seems to be stable but
Somnath ,
Yes , we are cloning repository in ceph client shared directory.
Please refer the time analyzation
Ceph Client Shared Directory :
[s...@cephclient2.csez.zohocorpin.com~/ceph/]$ time git clone
https://github.com/elastic/elasticsearch.git
Cloning into 'elasticsearch'...
On Tue, Jun 16, 2015 at 11:38 AM, Florent B flor...@coppint.com wrote:
I still have this problem on Hammer.
My CephFS directory contains 46MB of data, but the pool (configured with
layout, not default one) is 6.59GB...
How to debug this ?
On Mon, Mar 16, 2015 at 4:14 PM, John Spray
Hi all,
A cache tier, 2 hot node with 8 ssd osd, and 2 cold node with 24 sata osd.
The public network rate is 1Mb/s and cluster network rate is 1000Mb/s.
Using fuse-client to access the files.
The issue is:
When the files are in hot pool, the copy rate is very fash.
On Tue, Jun 16, 2015 at 11:53 AM, Jan Schermer j...@schermer.cz wrote:
Well, I see mons dropping out when deleting large amount of snapshots, and it
leats a _lot_ of CPU to delete them
Well, you're getting past my expertise on the subject, but deleting
snapshots can sometimes be expensive,
On 16 Jun 2015, at 12:59, Gregory Farnum g...@gregs42.com wrote:
On Tue, Jun 16, 2015 at 11:53 AM, Jan Schermer j...@schermer.cz wrote:
Well, I see mons dropping out when deleting large amount of snapshots, and
it leats a _lot_ of CPU to delete them
Well, you're getting past my
Ping :-)
Looks like nobody is bothered by this, can I assume it is normal, doesn’t hurt
anything and will grow to millions in time?
Jan
On 15 Jun 2015, at 10:32, Jan Schermer j...@schermer.cz wrote:
Hi,
I have ~1800 removed_snaps listed in the output of “ceph osd dump”.
Is that
Hello dear ceph developers and users,
I've spent some time tuning and measuring our ceph cluster performance,
and noticed quite strange thing..
I've been using fio (using both rbd engine on hosts and direct block (aio)
engine
inside qemu-kvm guests (qemu connected to ceph storage using rbd))
Thank you very much for your reply!
Is there anything I can do to go around that? e.g. setting access caps to
be released after a short while? Or is there a command to manually release
access caps (so that I could run it in cron)?
This is quite a problem because we have several applications that
On Tue, Jun 16, 2015 at 12:03 PM, Jan Schermer j...@schermer.cz wrote:
On 16 Jun 2015, at 12:59, Gregory Farnum g...@gregs42.com wrote:
On Tue, Jun 16, 2015 at 11:53 AM, Jan Schermer j...@schermer.cz wrote:
Well, I see mons dropping out when deleting large amount of snapshots, and
it leats
On Tue, Jun 16, 2015 at 3:30 AM, Jan Schermer j...@schermer.cz wrote:
Thanks for the answer.
So it doesn’t hurt performance if it grows to ridiculous size - e.g. no
lookup table overhead, stat()ing additional files etc.?
Nope, definitely nothing like that. If it gets sufficiently fragmented
Hello Cephers,
I have a question about pool quota. Is pool quota support RBD?
My cluster is Hammer 0.94.1 that have 1 Mon and 3 OSD. Each OSD server
have 3 disk.
My question is when I set pool quota size 1G on pool rbd. I still
can create a image abc = 3G.
After I mount and
Every time you delete a snapshot it goes in removed_snaps. The set of
removed snaps is stored as an interval set, so it uses up two integers
in the OSDMap for each range.
There are some patterns of usage that work out badly for this, but
generally if you're creating snapshots as time goes forward
On Mon, Jun 15, 2015 at 11:34 AM, negillen negillen negil...@gmail.com wrote:
Hello everyone,
something very strange is driving me crazy with CephFS (kernel driver).
I copy a large directory on the CephFS from one node. If I try to perform a
'time ls -alR' on that directory it gets executed
Thanks for the answer.
So it doesn’t hurt performance if it grows to ridiculous size - e.g. no lookup
table overhead, stat()ing additional files etc.?
Jan
On 16 Jun 2015, at 11:51, Gregory Farnum g...@gregs42.com wrote:
Every time you delete a snapshot it goes in removed_snaps. The set of
On Tue, Jun 16, 2015 at 12:11 PM, negillen negillen negil...@gmail.com wrote:
Thank you very much for your reply!
Is there anything I can do to go around that? e.g. setting access caps to be
released after a short while? Or is there a command to manually release
access caps (so that I could
Hi!
We also see this at our site: When we cat a large file from cephfs to
/dev/null, we get about 10MB/s data transfer. I also do not see a
system resource bottleneck.
Our cluster consists of 14 servers with each 16 disks, together forming
a EC coded pool. We also have 2SSDs per server for
On Tue, Jun 16, 2015 at 11:55 AM, Florent B flor...@coppint.com wrote:
On 06/16/2015 12:47 PM, Gregory Farnum wrote:
On Tue, Jun 16, 2015 at 11:38 AM, Florent B flor...@coppint.com wrote:
I still have this problem on Hammer.
My CephFS directory contains 46MB of data, but the pool (configured
Thanks again,
even 'du' performance is terrible on node B (testing on a directory taken
from Phoronix):
# time du -hs /storage/test9/installed-tests/pts/pgbench-1.5.1/
73M /storage/test9/installed-tests/pts/pgbench-1.5.1/
real0m21.044s
user0m0.010s
sys 0m0.067s
Reading the
You could use vmtouch to drop the specific directory from cache. It has the
ability to evict directories/files from memory.
From: negillen negillen negil...@gmail.com
To: Gregory Farnum g...@gregs42.com
Cc: ceph-users ceph-us...@ceph.com
Sent: Tuesday, June 16, 2015 7:37:23 AM
Subject: Re:
Have you tried just running “sync;sync” on the originating node? Does that
achieve the same thing or not? (I guess it could/should).
Jan
On 16 Jun 2015, at 13:37, negillen negillen negil...@gmail.com wrote:
Thanks again,
even 'du' performance is terrible on node B (testing on a
43 matches
Mail list logo