For readv/preadv sync-operatoin, ceph only do the first iov.
It don't think other iovs.Now implement this.
V5:
-before getattr,it must put caps which already holded avoid deadlock.
-only do generic_segment_checks for sync-read avoid do again in func
generic_file_aio_read
V4:
-modify one b
On 09/24/2013 04:05 PM, majianpeng wrote:
> diff --git a/fs/ceph/file.c b/fs/ceph/file.c
> index 3de8982..5422d8e 100644
> --- a/fs/ceph/file.c
> +++ b/fs/ceph/file.c
> @@ -408,51 +408,94 @@ more:
> *
> * If the read spans object boundary, just do multiple reads.
> */
> -static ssize_t ceph_s
>On 09/24/2013 04:05 PM, majianpeng wrote:
>> diff --git a/fs/ceph/file.c b/fs/ceph/file.c
>> index 3de8982..5422d8e 100644
>> --- a/fs/ceph/file.c
>> +++ b/fs/ceph/file.c
>> @@ -408,51 +408,94 @@ more:
>> *
>> * If the read spans object boundary, just do multiple reads.
>> */
>> -static ssiz
Hi,
Yes, the journal is a file on same XFS partition as the data.
I have some results from a rados bench to a single-copy pool on the JBOD disks.
[root@p05151113777233 ~]# rados bench -p test 30 write -t 1 -b 5
...
Total time run: 30.027296
Total writes made: 612
Write size:
On Mon, Sep 23, 2013 at 8:18 PM, Sage Weil wrote:
> You
> might try measuring that directly and comparing it to the 33ms
> append+fsync that you previously saw.
dd with fsync is quite slow...
[root@p05151113777233 fio]# time dd if=/dev/zero
of=/var/lib/ceph/osd/osd.1045/testtest bs=5 count=1
1+0
On 09/24/2013 04:54 PM, majianpeng wrote:
>> On 09/24/2013 04:05 PM, majianpeng wrote:
>>> diff --git a/fs/ceph/file.c b/fs/ceph/file.c
>>> index 3de8982..5422d8e 100644
>>> --- a/fs/ceph/file.c
>>> +++ b/fs/ceph/file.c
>>> @@ -408,51 +408,94 @@ more:
>>> *
>>> * If the read spans object bounda
>On 09/24/2013 04:54 PM, majianpeng wrote:
>>> On 09/24/2013 04:05 PM, majianpeng wrote:
diff --git a/fs/ceph/file.c b/fs/ceph/file.c
index 3de8982..5422d8e 100644
--- a/fs/ceph/file.c
+++ b/fs/ceph/file.c
@@ -408,51 +408,94 @@ more:
*
* If the read spans obje
>On 09/24/2013 04:05 PM, majianpeng wrote:
>> diff --git a/fs/ceph/file.c b/fs/ceph/file.c
>> index 3de8982..5422d8e 100644
>> --- a/fs/ceph/file.c
>> +++ b/fs/ceph/file.c
>> @@ -408,51 +408,94 @@ more:
>> *
>> * If the read spans object boundary, just do multiple reads.
>> */
>> -static ssiz
Some more tests...
~journal in XFS:
[root@p05151113777233 fio]# time dd if=/dev/zero
of=/var/lib/ceph/osd/osd.1045/testtest bs=512 count=1 conv=fsync
1+0 records in
1+0 records out
512 bytes (512 B) copied, 0.0423365 s, 12.1 kB/s
real 0m0.044s
user 0m0.000s
sys 0m0.001s
~directio via the FS
[ro
Ideally a partition using the first sectors of the disk.
I usually do a tiny partition at the beginning of the device and leave the rest
for ods_data.
Sébastien Han
Cloud Engineer
"Always give 100%. Unless you're giving blood.”
Phone: +33 (0)1 49 70 99 72
Mail: sebastien@enovance.com
A
Thanks Sebastien, we will try that. BTW, as you know we're using the
puppet modules of enovance, which set up a file-based journal. Did you
guys already puppetize a blockdev journal? I'll be doing that now if
not...
Cheers, Dan
CERN IT
On Tue, Sep 24, 2013 at 2:30 PM, Sébastien Han
wrote:
> Ideal
Hi guys,
Does anyone can point me to the piece of code called during the “ceph osd map
‘pool’ ‘object’” command please?
Thanks!
Sébastien Han
Cloud Engineer
"Always give 100%. Unless you're giving blood.”
Phone: +33 (0)1 49 70 99 72
Mail: sebastien@enovance.com
Address : 10, rue de la
Hi Dan,
Yes I noticed :), no we haven’t done anything on that side yet.
But I’ll be happy to see this happening, that way better than a file of the fs.
Would be nice if you could push something then :)
Cheers.
Sébastien Han
Cloud Engineer
"Always give 100%. Unless you're giving blood.”
Ph
On 09/24/2013 01:36 PM, Sébastien Han wrote:
Hi guys,
Does anyone can point me to the piece of code called during the “ceph osd map
‘pool’ ‘object’” command please?
Thanks!
$ git grep -n 'osd map' src/mon
src/mon/MonCommands.h:338:COMMAND("osd map " \
src/mon/OSDMonitor.cc:2093: } else if (p
Thanks Joao.
Sébastien Han
Cloud Engineer
"Always give 100%. Unless you're giving blood.”
Phone: +33 (0)1 49 70 99 72
Mail: sebastien@enovance.com
Address : 10, rue de la Victoire - 75009 Paris
Web : www.enovance.com - Twitter : @enovance
On September 24, 2013 at 3:07:52 PM, Joao Eduar
On Tue, 24 Sep 2013, Dan van der Ster wrote:
> Hi,
> Yes, the journal is a file on same XFS partition as the data.
>
> I have some results from a rados bench to a single-copy pool on the JBOD
> disks.
>
> [root@p05151113777233 ~]# rados bench -p test 30 write -t 1 -b 5
> ...
> Total time run:
On Tue, 24 Sep 2013, David Zafman wrote:
>
> Rados suite test run results for wip-5862. 2 "scrub mismatch" from mon
> (known problem). 2 are valgrind issues found with mds and osd.
What is the osd valgrind failure? And the osd.4 crash on 13443?
(Note that the teuthology.log will include mess
The osd.4 crash in 13443 is bug #5951:
2013-09-23 21:23:28.378428 1034e700 0 filestore(/var/lib/ceph/osd/ceph-4)
error (17) File exists not handled on operation 20 (6579.0.0, or op 0, counting
from 0)
2013-09-23 21:23:28.862204 1034e700 0 filestore(/var/lib/ceph/osd/ceph-4)
unexpected error
On Tue, 24 Sep 2013, Travis Rhoden wrote:
> This "noshare" option may have just helped me a ton -- I sure wish I would
> have asked similar questions sooner, because I have seen the same failure to
> scale. =)
>
> One question -- when using the "noshare" option (or really, even without it)
> are
On Tue, Sep 24, 2013 at 5:16 PM, Sage Weil wrote:
> On Tue, 24 Sep 2013, Travis Rhoden wrote:
>> This "noshare" option may have just helped me a ton -- I sure wish I would
>> have asked similar questions sooner, because I have seen the same failure to
>> scale. =)
>>
>> One question -- when using
Hi Somnath!
On Tue, 24 Sep 2013, Somnath Roy wrote:
>
> Hi Sage,
>
> We did quite a few experiment to see how ceph read performance can scale up.
> Here is the summary.
>
>
>
> 1.
>
> First we tried to see how far a single node cluster with one osd can scale
> up. We started with cuttlefish
See trace below. We run this command on system restart in order to
clear any blacklist which was created while node was mis-behaving.
Now, rather than giving a reasonable error, it causes at Traceback:
[root@node-172-20-0-13 ~]# ceph osd blacklist rm 172.20.0.13
Traceback (most recent call last):
Hi Sage,
Thanks for your input. I will try those. Please see my response inline.
Thanks & Regards
Somnath
-Original Message-
From: Sage Weil [mailto:s...@inktank.com]
Sent: Tuesday, September 24, 2013 3:47 PM
To: Somnath Roy
Cc: Travis Rhoden; Josh Durgin; ceph-devel@vger.kernel.org; Anir
For readv/preadv sync-operatoin, ceph only do the first iov.
It don't think other iovs.Now implement this.
V6:
Fix some bugs;
V5:
-before getattr,it must put caps which already holded avoid deadlock.
-only do generic_segment_checks for sync-read avoid do again in func
generic_file_aio_re
On 09/25/2013 09:07 AM, majianpeng wrote:
> For readv/preadv sync-operatoin, ceph only do the first iov.
> It don't think other iovs.Now implement this.
>
> V6:
>Fix some bugs;
> V5:
> -before getattr,it must put caps which already holded avoid deadlock.
> -only do generic_segment_checks f
>On 09/25/2013 09:07 AM, majianpeng wrote:
>> For readv/preadv sync-operatoin, ceph only do the first iov.
>> It don't think other iovs.Now implement this.
>>
>> V6:
>>Fix some bugs;
>> V5:
>> -before getattr,it must put caps which already holded avoid deadlock.
>> -only do generic_segment
On 09/25/2013 11:23 AM, majianpeng wrote:
>> On 09/25/2013 09:07 AM, majianpeng wrote:
>>> For readv/preadv sync-operatoin, ceph only do the first iov.
>>> It don't think other iovs.Now implement this.
>>>
>>> V6:
>>>Fix some bugs;
>>> V5:
>>> -before getattr,it must put caps which already ho
Applied this to the testing branch.
Thanks, everyone!
sage
On Wed, 25 Sep 2013, Yan, Zheng wrote:
> On 09/25/2013 11:23 AM, majianpeng wrote:
> >> On 09/25/2013 09:07 AM, majianpeng wrote:
> >>> For readv/preadv sync-operatoin, ceph only do the first iov.
> >>> It don't think other iovs.Now imp
28 matches
Mail list logo