Hi everyone,
We recently saw some extremely high stat loads on our lustre FS. Output from
“llstat -i 1 mdt” looked like:
[root@hpfs-fsl-mds0 lustre]#
/proc/fs/lustre/mds/MDS/mdt/stats @ 1530642446.366015124
Name Cur.Count Cur.Rate #Events Unit last
We use samba to export lustre from a linux lustre client to both Windows and
MacOS regularly. We are pretty much using samba defaults. But this is not
intended for serious use of lustre - mainly just for pulling a few files over
for presentations or to attach to an email, etc..
Yes, we experienced some similar slowness on our ZFS-based lustre FS too. More
details here:
http://lists.lustre.org/pipermail/lustre-discuss-lustre.org/2017-April/014390.html
This also affects git repos quite a bit. The fix suggested by Andreas in that
thread worked fairly well and we
We have a similar setup and recently had to do something similar - in our case,
to add a 2nd IB NID. The admin node says that servicenode is preferred over
failnode, so that's what we use. It works great - we love the capability to
fail over for maintenance or troubleshooting. Our
ess or implied, and I assume no
liability, etc. ☺
Nevertheless, I hope this helps, at least as a cross-reference.
Malcolm.
From: lustre-discuss <lustre-discuss-boun...@lists.lustre.org> on behalf of
"Vicker, Darby (JSC-EG311)" <darby.vicke...@nasa.gov>
Date: Saturday, 6 Janu
Sorry – one other question. We are configured for failover too. Will the "lctl
replace_nids" do the right thing or should I do the tunefs to make sure all the
failover pairs get updated properly? This is what our tunefs command would
look like for an OST:
tunefs.lustre \
Hello everyone,
We have an existing LFS that is dual-homed on ethernet (mainly for our
workstations) and IB (for the computational cluster), ZFS backend for the MDT
and OST's. We just got a new computational cluster and need to add another IB
NID. The procedure for doing this is straight
8.
Any suggestions?
Regards
Ravi Konila
From: Vicker, Darby (JSC-EG311)
Sent: Wednesday, October 25, 2017 11:51 PM
To: Mannthey, Keith ; Ravi Konila ; Lustre Discuss
Subject: Re: [lustre-discuss] MGS is not working in HA
Sorry – I also meant to say that the resolution went off the mailing list and
Sorry – I also meant to say that the resolution went off the mailing list and
was continued in LU-8397. You can find the patch there.
From: lustre-discuss on behalf of
Darby Vicker
Date: Wednesday, October 25, 2017 at 1:17 PM
Which version of lustre are you using? We initially has problem with this too
when using failover with lustre 2.8 and 2.9. We got a patch that fixed it and
recent versions work fine for us. We have a combined MGS/MDS so our scenario
is a little different but this sounds very similar to our
I'd recommend using this command to find all the files that were affected:
lfs find --ost > lost_files.txt
Then run an rsync to copy off the data that is left. Something like:
rsync --exclude lost_files.txt
From: lustre-discuss on behalf of
e MDT's for different file systems against the same MDS would
be as I have never attempted such a configuration.
-cf
On Fri, Sep 15, 2017 at 10:29 AM, Vicker, Darby (JSC-EG311)
<darby.vicke...@nasa.gov<mailto:darby.vicke...@nasa.gov>> wrote:
From: lustre-discuss
<lustre-discuss-bo
From: lustre-discuss on behalf of
Colin Faber
Date: Friday, September 15, 2017 at 9:48 AM
To: Ravi Konila
Cc: Lustre Discuss
Subject: Re: [lustre-discuss] Multiple Lustre
This sounds exactly like what we ran into when we upgraded to 2.9 (and is still
present in 2.10). See these:
http://lists.lustre.org/pipermail/lustre-discuss-lustre.org/2017-May/014524.html
https://jira.hpdd.intel.com/browse/LU-9574
The mailing list thread describes our problem a little more
Any more info on this? I’m running into the same thing. I tried to find and
LU on this but didn’t see anything directly related.
___
lustre-discuss mailing list
lustre-discuss@lists.lustre.org
Hello,
We've upgraded to 2.10 and I've been playing with progressive file layouts. To
begin, I'm just setting a test directory to use the following PFL.
lfs setstripe \
-E 4M -c 1 -S 1M -i -1 \
-E 256M -c 4 -S 1M -i -1 \
-E -1 -c 8 -S 4M -i -1 .
I then created some files in the
I’m not sure if there is something you can do on the server side for this but I
think you can remount read only on the clients fairly easily without disruption:
# mount -t lustre 192.x.x.x@tcp:/hpfs-fsl/ /mnt
# touch /mnt/test
# mount -t lustre -o remount,ro 192.x.x.x@tcp:/hpfs-fsl/ /mnt
# touch
Hello,
I've been looking through the LUG '17 docs and the references to the intel
manager for lustre caught my eye. I'd like to try it out but I'm having
trouble getting it compiled and haven't found much in the way of docs or
further information. If anyone knows where to find more info,
Using the git bisect we were able to isolate the problem to this commit:
commit d8467ab8a2ca15fbbd5be3429c9cf9ceb0fa78b8
LU-7990 clio: revise readahead to support 16MB IO
In our testing, we can read from a large file (stripe count=4) at near line
rate (10 GbE – so 1200 MB/s) using a client
>> I tried a 2.8 client mounting the 2.9 servers and that showed the expected
>> behavior ??? increasing
>> performance with increasing OST's. Two things:
>>
>> 1. Any pointers to compiling a 2.8 client on recent RHEL 7 kernels would be
>> helpful. I had to boot
>> into an older kernel
sa.gov>
Cc: "lustre-discuss@lists.lustre.org" <lustre-discuss@lists.lustre.org>
Subject: Re: [lustre-discuss] Large file read performance degradation from
multiple OST's
On May 24, 2017, at 10:04, Vicker, Darby (JSC-EG311) <darby.vicke...@nasa.gov>
wrote:
>
I tried a 2.8 client mounting the 2.9 servers and that showed the expected
behavior – increasing performance with increasing OST's. Two things:
1. Any pointers to compiling a 2.8 client on recent RHEL 7 kernels would be
helpful. I had to boot into an older kernel to get the above test done.
Hello,
We recently noticed that the large file read performance on our 2.9 LFS is
dramatically worse than it used to be. The attached plot is the result of a
test script that uses dd to write a large file (50GB) to disk, read that file
and then copy it to a 2nd file to test write, read and
This worked great. We implemented it on Friday and the timings of the dd test
on our 2.9/ZFS LFS have dropped to under a second. Thanks a lot. The risk of
both the client and OSS crashing within a few seconds is low enough for us
compared to the performance gain.
The commit you pointed to
Thanks for the kstat info. Our 2.4 LFS is quite a bit different architecture –
ldiskfs on a hardware RAID – so no opportunity to compare the zfs kstat info
between the two. Our 2.9 LFS is barely in production at this point and only a
handful of people have moved over to it. So its
Hello,
We are having a few performance issues with our newest lustre file system.
Here is the overview of our setup:
-) Supermicro servers connected to external 12Gb/s SAS JBODs for MDT/OSS storage
-) CentOS version = 7.3.1611 (kernel 3.10.0-514.2.2.el7.x86_64) on the servers
and clients
eck to see if you're experiencing any group upcall errors on the MDS?
Have you tried disabling group upcall on the MDS?
Also, when looking at the problematic files with lfs getstripe, have you
verified that those objects exist? Any errors about missing objects on the
OSTs? Any other errors?
On Mon,
, Murshid Azman
<murshid.az...@gmail.com>, "lustre-discuss@lists.lustre.org"
<lustre-discuss@lists.lustre.org>
Subject: Re: [lustre-discuss] Odd file permission help please. Has lustre
corrupted?
Is your authentication system across all of your lustre nodes in sync?
On Mon,
We have the same thing happening on one of our lustre file system (2.4.3
servers, 2.9 client). All OST’s are connected so that’s not the problem. We
do use ACL’s so I was hopeful this was the cause for us too. But this doesn’t
seem to be the case – any operation on the problem directory just
f.john...@aeoncomputing.com>> wrote:
You're also leaving out the corosync/pacemaker/stonith configuration. That is
unless you are doing manual export/import of pools.
On Fri, Feb 10, 2017 at 9:03 PM, Vicker, Darby (JSC-EG311)
<darby.vicke...@nasa.gov<mailto:darby.vicke...@nasa.gov>> wrote:
Sur
stand with OSTs and MDTs where all I really need is to have the
failnode set when I do the mkfs.lustre
However, as I understand it, you have to use something like pacemaker
and drbd to deal with the MGS/MGT. Is this how you approached it?
Brian Andrus
On 2/6/2017 12:58 PM, Vicker, Darb
Agreed. We are just about to go into production on our next LFS with the
setup described. We had to get past a bug in the MGS failover for
dual-homed servers but as of last week that is done and everything is
working great (see "MGS failover problem" thread on this mailing list from
this
I've gone back to having the tcp and o2ib NIDS again. Even though the ZFS
properties show all the NIDS, the info in /proc shows only a single NID. Its
no wonder the failover isn't working – it seems that the OST doesn't have any
failover NID's for the MGC like it did when I had only the tcp
Progress. I did another round of "tunefs.lustre –writeconf" to take out the IB
so we are on Ethernet only. I think the MDS/MGS failover worked properly –
note the "Connection restored to MGC192.52.98.30@tcp_1 (at 192.52.98.31@tcp)"
message in the oss logs below – and that the
>> Getting failover right over multiple separate networks can be a real
>> hair-pulling experience.
>
>Darby: Do you have the option of (at least temporarily) running the file
>system with only Infiniband configured? If you could set up the file system
>to only use Infiniband, then >that would
> The question I have in this is how long are you waiting, and how are you
> determining that lnet has hung?
The example I just sent today I waited about 10 minutes. But the other day it
looks like I waited about 20 minutes before rebooting as I couldn't kill lnet.
I'm calling it hung because
I tried a failover making sure lustre, including lnet, was completely shutdown
on the primary MDS. This didn't work either. Lnet hung like I remembered. So
I powered down the primary MDS to force it offline and then mounted lustre on
the secondary MDS. The services and a client recovers but
My understanding is that the MMP only works with ldiskfs – its not enabled with
a ZFS backend yet. But I could be wrong about that too. This is our first
attempt at setting up failover so we would like to get comfortable doing this
manually before we setup something automated.
I'll try the
node options if you wanted.
--
Rick Mohr
Senior HPC System Administrator
National Institute for Computational Sciences
http://www.nics.tennessee.edu
> On Jan 8, 2017, at 11:58 PM, Vicker, Darby (JSC-EG311)
<darby.vicke...@nasa.gov> wrote:
>
We have a new set of hardware we are configuring as a lustre file system. We
are having a problem with MGS failover and could use some help. It was
formatted originally using 2.8 but we have since upgraded to 2.9. We are using
a JBOB with server pairs for failover and are using ZFS as the
ystem this may take several hours.
>
>For disaster recovery purposes, a device-level backup (dd) is more "plug and
>play" in that the whole image is restored from the backup and the LFSCK phase
>only needs to handle files that have been modified since the time the backup
kup
was created.
Cheers, Andreas
--
Andreas Dilger
Lustre Principal Architect
Intel High Performance Data Division
On 2016/09/01, 14:49, "Vicker, Darby (JSC-EG311)" <darby.vicke...@nasa.gov>
wrote:
Thanks. This is happening on all the clients so its not a DLM lock problem.
first can also give you an idea of what kind of corruption
is present before Making the fix.
Cheers, Andreas
On Aug 31, 2016, at 10:54, Vicker, Darby (JSC-EG311)
<darby.vicke...@nasa.gov<mailto:darby.vicke...@nasa.gov>> wrote:
Hello,
We’ve run into a problem where an entire directo
Hello,
We’ve run into a problem where an entire directory on our lustre file system
has become inaccessible.
# mount | grep lustre2
192.52.98.142@tcp:/hpfs2eg3 on /lustre2 type lustre (rw,flock)
# ls -l /lustre2/mirrors/cpas
ls: cannot access /lustre2/mirrors/cpas: Stale file handle
# ls -l
44 matches
Mail list logo