[Gluster-users] can't use sqlite3 on gluster mounted as NFS

Craig Carl craig at gestas.net
Sat Sep 24 19:26:29 UTC 2011


Brandon -
    SQLite uses POSIX locking to implement some of its ACID compliant behavior and requires the filesystem to fully implement POSIX advisory locks. Most network filesystems (including Gluster native and NFS) don't support everything that SQLite needs and so using SQLite on a networked filesystem isn't recommend by the SQLite team, see this excerpt from the link I sent earlier -

SQLite uses POSIX advisory locks to implement locking on Unix. On Windows it uses the LockFile(), LockFileEx(), and UnlockFile() system calls. SQLite assumes that these system calls all work as advertised. If that is not the case, then database corruption can result. One should note that POSIX advisory locking is known to be buggy or even unimplemented on many NFS implementations (including recent versions of Mac OS X) and that there are reports of locking problems for network filesystems under Windows. Your best defense is to not use SQLite for files on a network filesystem.

Craig

Sent from a mobile device, please excuse my tpyos.

On Sep 24, 2011, at 0:19, Brandon Simmons <bsimmons at labarchives.com> wrote:

> On Fri, Sep 23, 2011 at 4:11 PM, Anand Babu Periasamy <ab at gluster.com> wrote:
>> This is a known issue. Gluster NFS doesn't support NLM (locking) yet. 3.4
>> may implement this.  Did you try on GlusterFS native mount?
> 
> Thanks for that information.
> 
> I did test with the native fuse mount, but the results were difficult
> to interpret. We have a rails application that writes to multiple
> sqlite databases, and a test script that simulates a bunch of random
> writes to a specified DB, retrying if it fails.
> 
> On NFS this test runs reasonably well: both clients take turns, a
> couple retries, all writes complete without failures.
> 
> But mounted over gluster (same machines, underlying disk as above) one
> client always runs while the other gets locked out (different client
> machines depending on which was started first). At some point during
> this test the client that was locked out from writing to the DB
> actually gets disconnected from gluster and I have to remount:
> 
>    $ ls /mnt/gluster
>    ls: cannot access /websites/: Transport endpoint is not connected
> 
> One client is consistently locked out even if they are writing to
> DIFFERENT DBs altogether.
> 
> The breakage of the mountpoint happened every time the test was run
> concurrently against the SAME DB, but did not seem to occur when
> clients were running against different DBs.
> 
> But like I said, this was a very high level test with many moving
> parts so I'm not sure how useful the above details are for you to
> know.
> 
> Happy to hear any ideas for testing,
> Brandon
> 
> /var/log/glusterfs/etc-glusterfs-glusterd.vol.log:
> [2011-09-16 19:32:38.122196] W
> [socket.c:1494:__socket_proto_state_machine] 0-socket.management:
> reading from socket failed. Error (Transport endpoint is not
> connected), peer (127.0.0.1:1017)
> 
>> 
>> --AB
>> 
>> On Sep 23, 2011 10:00 AM, "Brandon Simmons" <bsimmons at labarchives.com>
>> wrote:
>>> I am able to successfully mount a gluster volume using the NFS client
>>> on my test servers. Simple reading and writing seems to work, but
>>> trying to work with sqlite databases seems to cause the sqlite client
>>> and libraries to freeze. I have to send KILL to stop the process.
>>> 
>>> Here is an example, server 1 and 2 are clients mounting gluster volume
>>> over NFS:
>>> 
>>> server1# echo "working" > /mnt/gluster/test_simple
>>> server2# echo "working" >> /mnt/gluster/test_simple
>>> server1# cat /mnt/gluster/test_simple
>>> working
>>> working
>>> server1# sqlite3 /websites/new.sqlite3
>>> SQLite version 3.6.10
>>> Enter ".help" for instructions
>>> Enter SQL statements terminated with a ";"
>>> sqlite> create table memos(text, priority INTEGER);
>>> (...hangs forever, have to detach screen and do kill -9)
>>> 
>>> the gluster volume was created and NFS-mounted as per the instructions
>>> here:
>>> 
>>> 
>>> http://www.gluster.com/community/documentation/index.php/Gluster_3.2_Filesystem_Administration_Guide
>>> 
>>> If I mount the volume using the nolock option, then things work:
>>> 
>>> mount -t nfs -o nolock server:/test-vol /mnt/gluster
>>> 
>>> So I assume this has something to do with the locking RPC service
>>> stufff, which I don't know much about. Here's output from rpc info:
>>> 
>>> server# rpcinfo -p
>>> program vers proto port
>>> 100000 2 tcp 111 portmapper
>>> 100000 2 udp 111 portmapper
>>> 100024 1 udp 56286 status
>>> 100024 1 tcp 40356 status
>>> 100005 3 tcp 38465 mountd
>>> 100005 1 tcp 38466 mountd
>>> 100003 3 tcp 38467 nfs
>>> 
>>> 
>>> client1# rpcinfo -p server
>>> program vers proto port
>>> 100000 2 tcp 111 portmapper
>>> 100000 2 udp 111 portmapper
>>> 100024 1 udp 56286 status
>>> 100024 1 tcp 40356 status
>>> 100005 3 tcp 38465 mountd
>>> 100005 1 tcp 38466 mountd
>>> 100003 3 tcp 38467 nfs
>>> 
>>> client1# # rpcinfo -p
>>> program vers proto port
>>> 100000 2 tcp 111 portmapper
>>> 100000 2 udp 111 portmapper
>>> 100024 1 udp 32768 status
>>> 100024 1 tcp 58368 status
>>> 
>>> Thanks for any help,
>>> Brandon
>>> _______________________________________________
>>> Gluster-users mailing list
>>> Gluster-users at gluster.org
>>> http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
>> 
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20110924/710a2c63/attachment.html>


More information about the Gluster-users mailing list