I'm on the waiting list too, but anyway, not sure to come yet. (I come from
Lille, and I need to check with my boss to leave early)
By the way, any chance to have some ceph conference or meetup at solutions
linux expo at paris this year ?
- Mail original -
De: Loic Dachary
The documentation states that setting rbd cache = true in [global] enables
it, but doesn't elaborate on whether you need to restart any Ceph processe
It's on the client side ! (so no need to restart ceph daemons)
- Mail original -
De: Graeme Lambert glamb...@adepteo.net
À:
.conf :)
http://ceph.com/docs/next/rbd/qemu-rbd/
- Mail original -
De: Graeme Lambert glamb...@adepteo.net
À: Alexandre DERUMIER aderum...@odiso.com
Cc: ceph-users@lists.ceph.com
Envoyé: Jeudi 6 Février 2014 12:03:00
Objet: Re: [ceph-users] RBD Caching - How to enable?
Hi
] RBD Caching - How to enable?
Does anybody else think there is a problem with the docs/settings here...
Message: 13
Date: Thu, 06 Feb 2014 12:11:53 +0100 (CET)
From: Alexandre DERUMIER aderum...@odiso.com
To: Graeme Lambert glamb...@adepteo.net
Cc: ceph-users@lists.ceph.com
Subject
Hi,
I have some questions about comming cache pool feature.
Is it only a cache ? (are the datas on both cache pool and main pool ?)
Or are the datas migrated from the main pool to cache pool ?
Do we need to enable replication on the cache pool ?
What happen if we loose osds from cache pool ?
They were in french unfortunately.
Hi, I'm interested by the french presentations ;)
- Mail original -
De: Loic Dachary l...@dachary.org
À: Karan Singh karan.si...@csc.fi
Cc: ceph-users ceph-users@lists.ceph.com, Ceph Community
ceph-commun...@lists.ceph.com
Envoyé: Mardi 18 Février
Hi,
I'm looking to build a ceph cluster in a multi datacenter architecture.
Datas will be stored on DC1 and DC2 (mon + osds), and we'll have a third
datacenter DC3, with a monitor only for quorum
2x dark fiber
DC1-DC2
| |
|
Thanks for the report !
Results seem to be encouraging . (Is is leveldb keystore ?)
Thanks to fio-rbd, it'll be easier to do random io benchmark now !
(I'm waiting to see if rocksdb will improve things in the future)
Regards,
Alexandre
- Mail original -
De: Haomai Wang
Hi,
you should have -drive file=rbd:libvirt,discard=on in command line, to
have discard enabled
- Mail original -
De: ljm李嘉敏 jm...@ctrip.com
À: ceph-us...@ceph.com
Envoyé: Mercredi 5 Mars 2014 02:37:52
Objet: [ceph-users] Enabling discard/trim
Dear all,
I try to use
After apt-get update and upgrade i stil see 0.87 release .. any hint ?
what is your repository in sources.list ?
- Mail original -
De: Zeeshan Ali Shah zas...@pdc.kth.se
À: Florent MONTHEL fmont...@flox-arts.net
Cc: Sage Weil sw...@redhat.com, ceph-users ceph-us...@ceph.com, René
...@dachary.org
À: aderumier aderum...@odiso.com, ceph-users ceph-users@lists.ceph.com
Envoyé: Dimanche 1 Février 2015 18:42:51
Objet: Re: [ceph-users] erasure code : number of chunks for a small cluster ?
Hi Alexandre,
On 01/02/2015 18:15, Alexandre DERUMIER wrote:
Hi,
I'm currently trying
host can be down for maintenance.
Udo
PS: you also can't change k+m on a pool later...
On 01.02.2015 18:15, Alexandre DERUMIER wrote:
Hi,
I'm currently trying to understand how to setup correctly a pool with erasure
code
https://ceph.com/docs/v0.80/dev/osd_internals/erasure_coding
Hi,
From my tests with giant, this was the cpu which limit the performance on osd.
I'm going to do some benchmark with 2x10 cores 3,1ghz for 6ssd next month.
I'll post results on the mailing list.
- Mail original -
De: mad Engineer themadengin...@gmail.com
À: Gregory Farnum
debian deb packages update are not restarting services.
(So, I think it should be the same for ubuntu).
you need to restart daemons in this order:
-monitor
-osd
-mds
-rados gateway
http://ceph.com/docs/master/install/upgrading-ceph/
- Mail original -
De: Stephen Jahl
Oh, I didn't thinked about this.
Thanks Hector !
- Mail original -
De: Hector Martin hec...@marcansoft.com
À: ceph-users ceph-users@lists.ceph.com
Envoyé: Vendredi 6 Février 2015 09:06:29
Objet: Re: [ceph-users] erasure code : number of chunks for a small cluster ?
On 02/02/15 03:38,
Hi,
I'm currently use crush tunables optimal value.
If I upgrade from firefly to hammer, does the optimal value will upgrade to
optimal values for hammer.
So, does my clients (qemu-librbd) need to be also upgraded to hammer to support
new hammer features ?
If yes,
I think to:
- change
: optimal : upgrade from firefly to
hammer behaviour ?
On Mon, 9 Feb 2015, Alexandre DERUMIER wrote:
Hi,
I'm currently use crush tunables optimal value.
If I upgrade from firefly to hammer, does the optimal value will upgrade
to optimal values for hammer.
The tunables won't change
Hi,
From my last benchmark,
I was around 12 iops rand read 4k , 2iops rand write 4k (3 nodes with
2ssd osd+journal ssd intel 3500)
My main bottleneck was cpu (it's was 2x4cores 1,4ghz intel), both on osd and
client.
I'm going to test next month my production cluster, with bigger
...@gol.com
À: ceph-users ceph-users@lists.ceph.com
Cc: aderumier aderum...@odiso.com
Envoyé: Jeudi 22 Janvier 2015 09:37:03
Objet: Re: [ceph-users] Journals on all SSD cluster
Hello,
On Thu, 22 Jan 2015 08:32:13 +0100 (CET) Alexandre DERUMIER wrote:
Hi,
From my last benchmark,
Using which version
Hi,
you need to import foreign config from openmanage webui.
somewhere in storage controller
BTW, I'm currently testing new dell r630 with a perc h330 ( lsi 3008)
With this controller, it's possible to do hardware for some disks, and
passthrough for some others disks.
So, perfect for ceph
Nice Work Mark !
I don't see any tuning about sharding in the config file sample
(osd_op_num_threads_per_shard,osd_op_num_shards,...)
as you only use 1 ssd for the bench, I think it should improve results for
hammer ?
- Mail original -
De: Mark Nelson mnel...@redhat.com
À:
/2015 05:55, Alexandre DERUMIER wrote:
Hi,
I'm currently running a big mongodb cluster, around 2TB, (sharding +
replication).
And I have a lot of problems with mongo replication (out of syncs and need to
full replicate again and again datas between my mongo replicats).
So, I thinked
Hi,
Can you test with disabling rbd_cache ?
I remember of a bug detected in giant, not sure it's also the case for fireflt
This was this tracker:
http://tracker.ceph.com/issues/9513
But It has been solved and backported to firefly.
Also, can you test 0.80.6 and 0.80.7 ?
- Mail
Just buy it.
Nice book, I don't have read yet all the book, but it seem to cover all ceph
features.
Good job !
- Mail original -
De: Karan Singh karan.si...@csc.fi
À: Ceph Community ceph-commun...@lists.ceph.com, ceph-users
ceph-users@lists.ceph.com, ceph-maintain...@ceph.com,
Can this timeout be increased in some way? I've searched around and found the
/sys/block/sdx/device/timeout knob, which in my case is set to 30s.
yes, sure
echo 60 /sys/block/sdx/device/timeout
for 60s for example
- Mail original -
De: Krzysztof Nowicki
Hi,
Intel has just released new ssd s3610:
http://www.anandtech.com/show/8954/intel-launches-ssd-dc-s3610-s3710-enterprise-ssds
endurance is 10x bigger than 3500, for 10% cost addition.
Has somebody already tested them ?
Regards,
Alexandre
___
Hi,
I'm currently trying to understand how to setup correctly a pool with erasure
code
https://ceph.com/docs/v0.80/dev/osd_internals/erasure_coding/developer_notes/
My cluster is 3 nodes with 6 osd for each node (18 osd total).
I want to be able to survive of 2 disk failures, but also a
Hi,
also check your cpu usage, dell poweredge 2900 are quite old (6-8 years old),
The more iops you need, to more cpu you need.
I don't remember what is the default blocksize of rados bench.
- Mail original -
De: Patrik Plank pat...@plank.me
À: ceph-users ceph-users@lists.ceph.com
Hi,
Do you have also tested 0.80.6 and 0.80.7 librbd ?
could be usefull to search commits in git.
(I'm not sure that all changes are in the release note)
- Mail original -
De: 杨万元 yangwanyuan8...@gmail.com
À: ceph-users ceph-users@lists.ceph.com
Envoyé: Jeudi 12 Février 2015 04:14:15
To my surprise however these slow requests caused aborts from the block
device on the VM side, which ended up corrupting files
This is very strange, you shouldn't have corruption.
Do you use writeback ? if yes, do you have disable barrier on your filesystem ?
(What is the qemu version ? guest
Hi,
I'm currently running a big mongodb cluster, around 2TB, (sharding +
replication).
And I have a lot of problems with mongo replication (out of syncs and need to
full replicate again and again datas between my mongo replicats).
So, I thinked to use rbd to replicate the storage and keep
Hi,
What I saw after enabling RBD cache it is working as expected, means
sequential write has better MBps than random write. can somebody explain this
behaviour ?
This is because rbd_cache merge coalesced ios in bigger ios, so it's working
only with sequential workload.
you'll do less ios
will see it when the stuff is being
assembled :o
Do you feel that there something in the setup that could be better if you would
decide on hardware as of today?
Also, will you try out replication set of 2 as well?
Thanks
Josef
On 18 Mar 2015, at 08:19, Alexandre DERUMIER aderum...@odiso.com
Hi,
I don't known how rbd read-ahead is working,
but with qemu virtio-scsi, you can have read merge request (for sequential
reads), so it's doing bigger ops to ceph cluster and improve throughput.
virtio-blk merge request will be supported in coming qemu 2.3.
(I'm not sure of virtio-win
Hi, you don't need to defined ip and host for osd,
but you need to defined monitor ips
[osd]
keyring = /var/lib/ceph/osd/ceph-$id/keyring
[mon.0]
host = node1
mon addr = X.X.X.X:6789
[mon.1]
host = node2
mon addr = X.X.X.X:6789
[mon.2]
host
Hi Josef,
I'm going to benchmark a 3nodes cluster with 6ssd each node (2x10 cores 3,1ghz).
From my previous bench, you need fast cpus if you need a lot of iops, and
writes are lot more expansive than reads.
Now i'm you are doing only small iops (big blocks / big throughput), you don't
need too
That full system slows down, OK, but brutal stop...
This is strange, that could be:
- qemu crash, maybe a bug in rbd block storage (if you use librbd)
- oom-killer on you host (any logs ?)
what is your qemu version ?
- Mail original -
De: Florent Bautista flor...@coppint.com
À:
: aderumier aderum...@odiso.com
Cc: ceph-users ceph-users@lists.ceph.com
Envoyé: Lundi 16 Mars 2015 11:14:45
Objet: Re: [ceph-users] [SPAM] Changing pg_num = RBD VM down !
On 03/16/2015 11:03 AM, Alexandre DERUMIER wrote:
This is strange, that could be:
- qemu crash, maybe a bug in rbd block
Hi,
I'll build my full ssd production soon,
I wonder which distrib is best tested with inktank and ceph team ?
ceph.com doc is quite old, and don't have reference for giant or hammer
http://ceph.com/docs/master/start/os-recommendations/
Seem than in past only ubuntu and rhel was well tested,
aderum...@odiso.com
Cc: ceph-users ceph-users@lists.ceph.com
Envoyé: Lundi 16 Mars 2015 12:35:11
Objet: Re: [ceph-users] [SPAM] Changing pg_num = RBD VM down !
On 03/16/2015 12:23 PM, Alexandre DERUMIER wrote:
We use Proxmox, so I think it uses librbd ?
As It's me that I made the proxmox rbd plugin
Is it possible all replicas of an object to be saved in the same node?
No. (until you don't wrongly modify the crushmap manually)
Is it possible to lose any?
with replicat x2, if you loose 2osd on 2differents nodes, with the same object
inside, you'll lost the object
Is there a mechanism
: 06 March 2015 16:02
To: Alexandre DERUMIER; Nick Fisk
Cc: ceph-users
Subject: Re: [ceph-users] Strange krbd behaviour with queue depths
Nick,
I think this is because of the krbd you are using is using Naggle's
algorithm i.e
TCP_NODELAY = false by default.
The latest krbd module
I think this will be fixed in next firefly point release
tracker for firefly 0.80.8 speed decrease
http://tracker.ceph.com/issues/10956
Jason Dillaman link it to the famous object_cacher bug:
http://tracker.ceph.com/issues/9854
- Mail original -
De: Gregory Farnum g...@gregs42.com
À:
cluster with 3 nodes and 9 OSD
with 3.16-3 kernel
Am 28.02.2015 um 12:43 schrieb Alexandre DERUMIER aderum...@odiso.com:
Hi,
First, test if your ssd can write fast with O_DSYNC
check this blog:
http://www.sebastien-han.fr/blog/2014/10/10/ceph-how-to-test-if-your-ssd-is-suitable
As optimisation,
try to set ioscheduler to noop,
and also enable rbd_cache=true. (It's really helping for for sequential writes)
but your results seem quite low, 926kb/s with 4k, it's only 200io/s.
check if you don't have any big network latencies, or mtu fragementation
problem.
Maybe also
Hi,
First, test if your ssd can write fast with O_DSYNC
check this blog:
http://www.sebastien-han.fr/blog/2014/10/10/ceph-how-to-test-if-your-ssd-is-suitable-as-a-journal-device/
Then, try with ceph Giant (or maybe wait for Hammer), because they are a lot of
optimisations for ssd for threads
Hi,
from qemu rbd.c
if (flags BDRV_O_NOCACHE) {
rados_conf_set(s-cluster, rbd_cache, false);
} else {
rados_conf_set(s-cluster, rbd_cache, true);
}
and
block.c
int bdrv_parse_cache_flags(const char *mode, int *flags)
{
*flags = ~BDRV_O_CACHE_MASK;
if
I just upgraded my debian giant cluster,
1)on each node:
-
apt-get update
apt-get dist-upgrade
2)on each node:
-
/etc/init.d/ceph restart mon
#ceph -w --- verify that HEALTH is ok before doing another node
3)on each node:
-
/etc/init.d/ceph restart osd
comparison
thread, Alexandre DERUMIER wondered if changing the default shard and
threads per shard OSD settings might have a positive effect on
performance in our tests. I went back and used one of the PCIe SSDs
from our previous tests to experiment with a recent master pull. I
wanted to know how
Mars 2015 16:42:13
Objet: Re: [ceph-users] Perf problem after upgrade from dumpling to firefly
Only writes ;)
Le mercredi 04 mars 2015 à 16:19 +0100, Alexandre DERUMIER a écrit :
The change is only on OSD (and not on OSD journal).
do you see twice iops for read and write ?
if only read
Hi for hardware, inktank have good guides here:
http://www.inktank.com/resource/inktank-hardware-selection-guide/
http://www.inktank.com/resource/inktank-hardware-configuration-guide/
ceph works well with multiple osd daemon (1 osd by disk),
so you should not use raid.
(xfs is the recommended
(and not on OSD journal).
Le mercredi 04 mars 2015 à 15:05 +0100, Alexandre DERUMIER a écrit :
The load problem is permanent : I have twice IO/s on HDD since firefly.
Oh, permanent, that's strange. (If you don't see more traffic coming from
clients, I don't understand...)
do you see also
Hi,
maybe this is related ?:
http://tracker.ceph.com/issues/9503
Dumpling: removing many snapshots in a short time makes OSDs go berserk
http://tracker.ceph.com/issues/9487
dumpling: snaptrimmer causes slow requests while backfilling.
osd_snap_trim_sleep not helping
firefly.
And yes, the problem hang the production at night during snap trimming.
I suppose there is a new OSD parameter which change behavior of the
journal, or something like that. But didn't find anything about that.
Olivier
Le mercredi 04 mars 2015 à 14:44 +0100, Alexandre DERUMIER
,
In the Ceph Dumpling/Firefly/Hammer SSD/Memstore performance comparison
thread, Alexandre DERUMIER wondered if changing the default shard and
threads per shard OSD settings might have a positive effect on
performance in our tests. I went back and used one of the PCIe SSDs
from our previous tests
2015 22:49:23
Objet: Re: [ceph-users] Ceph Hammer OSD Shard Tuning Test Results
Can I ask what xio and simple messenger are and the differences?
Kind regards
Kevin Walker
+968 9765 1742
On 1 Mar 2015, at 18:38, Alexandre DERUMIER aderum...@odiso.com wrote:
Hi Mark,
I found an previous
can sure this problem is cause from 0.80.8
2015-02-12 19:33 GMT+08:00 Alexandre DERUMIER aderum...@odiso.com :
Hi,
Can you test with disabling rbd_cache ?
I remember of a bug detected in giant, not sure it's also the case for
fireflt
This was this tracker:
http://tracker.ceph.com/issues
Hi,
I known that Loic Dachary was currently working on backporting new feature on
giant,
I see that 0.87.1 has been tagged in git too:
here the difference:
https://github.com/ceph/ceph/compare/v0.87...v0.87.1
Loic, any annoucement/release note, yet ?
- Mail original -
De: Lindsay
Hi,
dd if=/dev/zero of=tempfile bs=1M count=1024 conv=fdatasync,notrunc
1073741824 Bytes (1,1 GB) kopiert, 2,53986 s, 423 MB/s
How much do you get with o_dsync? (ceph journal use o_dsync, and some ssd are
pretty slow with dsync)
07:56:33 +0100 (CET) Alexandre DERUMIER wrote:
Hi,
dd if=/dev/zero of=tempfile bs=1M count=1024 conv=fdatasync,notrunc
1073741824 Bytes (1,1 GB) kopiert, 2,53986 s, 423 MB/s
How much do you get with o_dsync? (ceph journal use o_dsync, and some
ssd are pretty slow with dsync
Hi,
Isn't it in the nature of ssd to have write amplication ?
Generaly, they have a erase block size of 128k,
so the worst case could be 128/4 = 32x write amplification.
(of course ssd algorithms and optimisations reduce this write amplification).
Now, it could be great to see if it's coming
(not tested, but I think with journal and O_DSYNC writes, it can give use ssd
write amplification)
also, I think that enterprise ssd with supercapacitor, should be able to cache
theses o_dsync writes in
the ssd buffer, and do bigger writes to reduce amplification.
Don't known how ssd internal
performance seem quite impressive (40iops 4k) :)
for endurance, It's looking like intel s3500
I still preferer s3610 for now,
400GB for $419 , with 3.0PB endurance.
- Mail original -
De: Mark Nelson mnel...@redhat.com
À: ceph-users ceph-users@lists.ceph.com
Envoyé: Jeudi 2
Hi,
for writes, ceph write twice to the disk, 1 for journal 1 for datas. (so
half write bandwith)
and journal is writen with O_DSYNC
(you should test your disk with fio --sync=1 to compare).
That's why the recommandation is to use ssd for journal disks.
- Mail original -
De:
Hi,
I'm currently benching full ssd setup (don't have finished yet),
but with 4osd, ssd intel s3500, (replication x1), with randwrite 4M, I'm around
550MB/S
with random 4K, i'm around 4iops (1iops by osd, limit is the disk
write o_dsync speed)
This is with hammer.
- Mail
[mailto:ceph-users-boun...@lists.ceph.com] On Behalf Of Mark
Nelson
Sent: Wednesday, April 22, 2015 7:31 PM
To: Alexandre DERUMIER; Milosz Tanski
Cc: ceph-devel; ceph-users
Subject: Re: [ceph-users] strange benchmark problem : restarting osd daemon
improve performance from 100k iops to 300k iops
: this is one of the reasons I like small nodes with single sockets
and fewer OSDs.
Mark
On 04/22/2015 08:56 AM, Alexandre DERUMIER wrote:
Hi,
I have done a lot of test today, and it seem indeed numa related.
My numastat was
# numastat
node0 node1
numa_hit 99075422 153976877
numa_miss
from 100k iops to 300k iops
On Wed, Apr 22, 2015 at 5:01 AM, Alexandre DERUMIER aderum...@odiso.com
wrote:
I wonder if it could be numa related,
I'm using centos 7.1,
and auto numa balacning is enabled
cat /proc/sys/kernel/numa_balancing = 1
Maybe osd daemon access to buffer on wrong
it a shot. :D
Mark
On 04/24/2015 12:36 PM, Stefan Priebe - Profihost AG wrote:
Is jemalloc recommanded in general? Does it also work for firefly?
Stefan
Excuse my typo sent from my mobile phone.
Am 24.04.2015 um 18:38 schrieb Alexandre DERUMIER aderum...@odiso.com
mailto:aderum
.
Thanks Regards
Somnath
-Original Message-
From: ceph-users [mailto: ceph-users-boun...@lists.ceph.com ] On Behalf Of
Alexandre DERUMIER
Sent: Thursday, April 23, 2015 4:56 AM
To: Mark Nelson
Cc: ceph-users; ceph-devel; Milosz Tanski
Subject: Re: [ceph-users] strange benchmark problem
I'm able to reach around 2-25000iops with 4k block with s3500 (with
o_dsync) (so yes, around 80-100MB/S).
I'l bench new s3610 soon to compare.
- Mail original -
De: Anthony Levesque aleves...@gtcomm.net
À: Christian Balzer ch...@gol.com
Cc: ceph-users ceph-users@lists.ceph.com
,
Srinivas
-Original Message-
From: ceph-users [mailto:ceph-users-boun...@lists.ceph.com] On Behalf Of Mark
Nelson
Sent: Wednesday, April 22, 2015 7:31 PM
To: Alexandre DERUMIER; Milosz Tanski
Cc: ceph-devel; ceph-users
Subject: Re: [ceph-users] strange benchmark problem
[mailto:ceph-users-boun...@lists.ceph.com] On Behalf Of Mark
Nelson
Sent: Wednesday, April 22, 2015 7:31 PM
To: Alexandre DERUMIER; Milosz Tanski
Cc: ceph-devel; ceph-users
Subject: Re: [ceph-users] strange benchmark problem : restarting osd daemon
improve performance from 100k iops to 300k
how it does.
In some ways I'm glad it turned out not to be NUMA. I still suspect we
will have to deal with it at some point, but perhaps not today. ;)
Mark
On 04/23/2015 05:58 AM, Alexandre DERUMIER wrote:
Maybe it's tcmalloc related
I thinked to have patched it correctly, but perf show
Hi,
I was doing some benchmarks,
I have found an strange behaviour.
Using fio with rbd engine, I was able to reach around 100k iops.
(osd datas in linux buffer, iostat show 0% disk access)
then after restarting all osd daemons,
the same fio benchmark show now around 300k iops.
(osd datas in
I wonder if it could be numa related,
I'm using centos 7.1,
and auto numa balacning is enabled
cat /proc/sys/kernel/numa_balancing = 1
Maybe osd daemon access to buffer on wrong numa node.
I'll try to reproduce the problem
- Mail original -
De: aderumier aderum...@odiso.com
À:
)
Thanks in advance,
JV
On Sun, Apr 26, 2015 at 10:46 PM, Alexandre DERUMIER
aderum...@odiso.com wrote:
I'll retest tcmalloc, because I was prety sure to have patched it correctly.
Ok, I really think I have patched tcmalloc wrongly.
I have repatched it, reinstalled it, and now I'm getting
it a shot. :D
Mark
On 04/24/2015 12:36 PM, Stefan Priebe - Profihost AG wrote:
Is jemalloc recommanded in general? Does it also work for firefly?
Stefan
Excuse my typo sent from my mobile phone.
Am 24.04.2015 um 18:38 schrieb Alexandre DERUMIER aderum...@odiso.com
mailto:aderum
Alexandre DERUMIER aderum...@odiso.com
mailto:aderum...@odiso.com:
Hi,
I have finished to rebuild ceph with jemalloc,
all seem to working fine.
I got a constant 300k iops for the moment, so no speed regression.
I'll do more long benchmark next week.
Regards,
Alexandre
Hi, can you check on your ceph node
/var/log/salt/minion
?
I have had some similar problem, I have need to remove
rm /etc/salt/pki/minion/minion_master.pub
/etc/init.d/salt-minion restart
(I don't known if calamari-ctl clear change the salt master key)
- Mail original -
De:
conclusions about
jemalloc vs tcmalloc until we can figure out what went wrong.
Mark
On 04/27/2015 12:46 AM, Alexandre DERUMIER wrote:
I'll retest tcmalloc, because I was prety sure to have patched it
correctly.
Ok, I really think I have patched tcmalloc wrongly.
I have repatched
Avril 2015 18:34:50
Objet: Re: [ceph-users] strange benchmark problem : restarting osd daemon
improve performance from 100k iops to 300k iops
On 04/27/2015 10:11 AM, Alexandre DERUMIER wrote:
Is it possible that you were suffering from the bug during the first
test but once reinstalled you
maybe this could help to repair pgs ?
http://www.sebastien-han.fr/blog/2015/04/27/ceph-manually-repair-object/
(6 disk at the same time seem pretty strange. do you have some kind of
writeback cache enable of theses disks ?)
- Mail original -
De: Yujian Peng
for the moment, you can use snapshot for backup
https://ceph.com/community/blog/tag/backup/
I think that async mirror is on the roadmap
https://wiki.ceph.com/Planning/Blueprints/Hammer/RBD%3A_Mirroring
if you use qemu, you can do qemu full backup. (qemu incremental backup is
coming for qemu
Hi,
If I have two networks (public and cluster network) and one link in public
network is broken ( cluster network is fine) what I will see in my cluster ?
See
http://ceph.com/docs/master/rados/configuration/network-config-ref/
only osd between them use private network.
so if public
I tries searching on internet and could not find a el7 package with
liburcu-bp.la file, let me know which rpm package has this libtool archive.
Hi, maybe can you try
./install-deps.sh
to install needed dependencies.
- Mail original -
De: Srikanth Madugundi
Hi,
rbd_cache is client config only,
so no need to restart osd.
if you set cache=writeback in libvirt, it'll enable it,
so you don't need to setup rbd_cache=true in ceph.conf.
(it should override it)
you can verify it enable, doing a sequantial write benchmark with 4k block.
you should have a
Hi,
as Debian Jessie is already released for some time, I'd like to ask is
there any plans to build newer Ceph packages for it?
Yes it's planned, I'm currently helping on create images for build
integration platform.
If you want, I have build ceph packages for jessie:
Hi, when you have remove salt from nodes,
do you have remove the old master key
/etc/salt/pki/minion/minion_master.pub
?
I have add the same behavior than you when reinstalling calamari server, and
previously installed salt on ceph nodes
(with explicit error about the key in
Hi,
I'm currently doing benchmark too, and I don't see this behavior
I get very nice performance of up to 200k IOPS. However once the volume is
written to (ie when I map it using rbd map and dd whole volume with some
random data),
and repeat the benchmark, random performance drops to ~23k IOPS.
Hi,
I'm currently try to deploy a new ceph test cluster on centos7, (hammer)
from ceph-deploy (on a debian wheezy).
And it seem that systemd unit files are not deployed
Seem that ceph git have systemd unit file
https://github.com/ceph/ceph/tree/hammer/systemd
I don't have look inside the rpm
So what would you suggest, what are your experiences?
Hi, you can have a look at mellanox sx1012 for example
http://www.mellanox.com/page/products_dyn?product_family=163
12 ports 40GB for around 4000€
you can use breakout cables to have 4x12 10GB ports.
They can be stacked with mlag and lacp
Hi Mark,
I finally got my hardware for my production full ssd cluster.
Here a first preliminary bench. (1osd).
I got around 45K iops with randread 4K with a small 10GB rbd volume
I'm pretty happy because I don't see anymore huge cpu difference between krbd
lirbd.
In my previous bench I was
Kozanecki | Linux Administrator | E: mkozane...@evertz.com
-Original Message-
From: ceph-users [mailto:ceph-users-boun...@lists.ceph.com] On Behalf Of
Alexandre DERUMIER
Sent: April-17-15 11:38 AM
To: Mark Nelson; ceph-users
Subject: [ceph-users] full ssd setup preliminary hammer
...@redhat.com,
ceph-users ceph-users@lists.ceph.com
Envoyé: Vendredi 17 Avril 2015 20:57:42
Objet: Re: [ceph-users] full ssd setup preliminary hammer bench
Am 17.04.2015 um 17:37 schrieb Alexandre DERUMIER:
Hi Mark,
I finally got my hardware for my production full ssd cluster.
Here a first
interfaces per node. Does that not limit
the thgouthput and hence the load on the CPUs?
-Sreenath
On 4/2/15, Alexandre DERUMIER aderum...@odiso.com wrote:
Hi,
with HP SL4540 server?
this model
http://www8.hp.com/h20195/v2/GetDocument.aspx?docname=c04128155
?
(60
: [ceph-users] full ssd setup preliminary hammer bench
Am 18.04.2015 um 07:24 schrieb Alexandre DERUMIER aderum...@odiso.com:
any idea whether this might be the tcmalloc bug?
I still don't known if centos/redhat packages have also the bug or not.
gperftools.x86_64 2.1-1.el7
From
Hi,
looking at the latest version of QEMU,
It's seem that it's was already this behaviour since the add of rbd_cache
parsing in rbd.c by josh in 2012
-users] rbd cache + libvirt
On Tue, Jun 9, 2015 at 7:59 AM, Alexandre DERUMIER aderum...@odiso.com wrote:
host conf : rbd_cache=true : guest cache=none : result : cache (wrong)
Thanks Alexandre, so you are confirming that this exact case misbehaves
to performance.
It would nice to route cause the problem if that is the case.
On Tue, Jun 9, 2015 at 11:21 AM, Alexandre DERUMIER aderum...@odiso.com
wrote:
Hi,
I'm doing benchmark (ceph master branch), with randread 4k qdepth=32,
and rbd_cache=true seem to limit the iops around 40k
no cache
101 - 200 of 401 matches
Mail list logo