Discussion:
Bug 1374166 or similar
Add Reply
Bernhard Dübi
2017-07-14 07:36:03 UTC
Reply
Permalink
Raw Message
Hello everybody,

I'm in a similar situation as described in
https://bugzilla.redhat.com/show_bug.cgi?id=1374166


I have a gluster volume exported through ganesha. we had some problems
on the gluster server and the NFS mount on the client was hanging.
I did a lazy umount of the NFS mount on the client, then went to the
Gluster server, mounted the Gluster volume and deleted a bunch of
files.
When I mounted the volume again on the client I noticed that the space
was not freed. Now I find them in $brick/.glusterfs/unlink

OS: Ubuntu 16.04
Gluster: 3.8.13
Ganesha: 2.4.5

Let me know if you need more info

Best Regards
Bernhard
Jiffin Tony Thottan
2017-07-14 08:43:34 UTC
Reply
Permalink
Raw Message
Post by Bernhard Dübi
Hello everybody,
I'm in a similar situation as described in
https://bugzilla.redhat.com/show_bug.cgi?id=1374166
The issue got fixed by https://review.gluster.org/#/c/14820 and is
already available in 3.8 branch
Post by Bernhard Dübi
I have a gluster volume exported through ganesha. we had some problems
on the gluster server and the NFS mount on the client was hanging.
I did a lazy umount of the NFS mount on the client, then went to the
Gluster server, mounted the Gluster volume and deleted a bunch of
files.
When I mounted the volume again on the client I noticed that the space
was not freed. Now I find them in $brick/.glusterfs/unlink
Here you have mounted the volume via glusterfs fuse mount and deleted
those files
right(not directly from the bricks)?
Can you restart nfs-ganesha server and see what happens ?
What type of volume are you using?
--
Jiffin
Post by Bernhard Dübi
OS: Ubuntu 16.04
Gluster: 3.8.13
Ganesha: 2.4.5
Let me know if you need more info
Best Regards
Bernhard
_______________________________________________
Gluster-users mailing list
http://lists.gluster.org/mailman/listinfo/gluster-users
Bernhard Dübi
2017-07-14 10:43:28 UTC
Reply
Permalink
Raw Message
Hi,

yes, I mounted the Gluster volume and deleted the files from the
volume not the brick

mount -t glusterfs hostname:volname /mnt
cd /mnt/some/directory
rm -rf *

restart of nfs-ganesha is planned for tomorrow. I'll keep you posted
BTW: nfs-ganesha is running on a separate server in standalone configuration

Best Regards
Bernhard
Post by Bernhard Dübi
Hello everybody,
I'm in a similar situation as described in
https://bugzilla.redhat.com/show_bug.cgi?id=1374166
The issue got fixed by https://review.gluster.org/#/c/14820 and is already
available in 3.8 branch
Post by Bernhard Dübi
I have a gluster volume exported through ganesha. we had some problems
on the gluster server and the NFS mount on the client was hanging.
I did a lazy umount of the NFS mount on the client, then went to the
Gluster server, mounted the Gluster volume and deleted a bunch of
files.
When I mounted the volume again on the client I noticed that the space
was not freed. Now I find them in $brick/.glusterfs/unlink
Here you have mounted the volume via glusterfs fuse mount and deleted those
files
right(not directly from the bricks)?
Can you restart nfs-ganesha server and see what happens ?
What type of volume are you using?
--
Jiffin
Post by Bernhard Dübi
OS: Ubuntu 16.04
Gluster: 3.8.13
Ganesha: 2.4.5
Let me know if you need more info
Best Regards
Bernhard
_______________________________________________
Gluster-users mailing list
http://lists.gluster.org/mailman/listinfo/gluster-users
Bernhard Dübi
2017-07-16 14:41:14 UTC
Reply
Permalink
Raw Message
Hi,

both Gluster servers were rebooted and now the unlink directory is clean.

Best Regards
Bernhard
Post by Bernhard Dübi
Hi,
yes, I mounted the Gluster volume and deleted the files from the
volume not the brick
mount -t glusterfs hostname:volname /mnt
cd /mnt/some/directory
rm -rf *
restart of nfs-ganesha is planned for tomorrow. I'll keep you posted
BTW: nfs-ganesha is running on a separate server in standalone configuration
Best Regards
Bernhard
Post by Bernhard Dübi
Hello everybody,
I'm in a similar situation as described in
https://bugzilla.redhat.com/show_bug.cgi?id=1374166
The issue got fixed by https://review.gluster.org/#/c/14820 and is already
available in 3.8 branch
Post by Bernhard Dübi
I have a gluster volume exported through ganesha. we had some problems
on the gluster server and the NFS mount on the client was hanging.
I did a lazy umount of the NFS mount on the client, then went to the
Gluster server, mounted the Gluster volume and deleted a bunch of
files.
When I mounted the volume again on the client I noticed that the space
was not freed. Now I find them in $brick/.glusterfs/unlink
Here you have mounted the volume via glusterfs fuse mount and deleted those
files
right(not directly from the bricks)?
Can you restart nfs-ganesha server and see what happens ?
What type of volume are you using?
--
Jiffin
Post by Bernhard Dübi
OS: Ubuntu 16.04
Gluster: 3.8.13
Ganesha: 2.4.5
Let me know if you need more info
Best Regards
Bernhard
_______________________________________________
Gluster-users mailing list
http://lists.gluster.org/mailman/listinfo/gluster-users
Jiffin Tony Thottan
2017-07-18 06:53:49 UTC
Reply
Permalink
Raw Message
Post by Bernhard Dübi
Hi,
both Gluster servers were rebooted and now the unlink directory is clean.
Following should have happened, If delete operation is performed gluster
keeps file in .unlink directory if it has open fd.
In this case since lazy umount is performed, ganesha server may still
keep the fd's open by that client so gluster keeps
the unlink directory even though it is removed from fuse mount.

--
Jiffin
Post by Bernhard Dübi
Best Regards
Bernhard
Post by Bernhard Dübi
Hi,
yes, I mounted the Gluster volume and deleted the files from the
volume not the brick
mount -t glusterfs hostname:volname /mnt
cd /mnt/some/directory
rm -rf *
restart of nfs-ganesha is planned for tomorrow. I'll keep you posted
BTW: nfs-ganesha is running on a separate server in standalone configuration
Best Regards
Bernhard
Post by Bernhard Dübi
Hello everybody,
I'm in a similar situation as described in
https://bugzilla.redhat.com/show_bug.cgi?id=1374166
The issue got fixed by https://review.gluster.org/#/c/14820 and is already
available in 3.8 branch
Post by Bernhard Dübi
I have a gluster volume exported through ganesha. we had some problems
on the gluster server and the NFS mount on the client was hanging.
I did a lazy umount of the NFS mount on the client, then went to the
Gluster server, mounted the Gluster volume and deleted a bunch of
files.
When I mounted the volume again on the client I noticed that the space
was not freed. Now I find them in $brick/.glusterfs/unlink
Here you have mounted the volume via glusterfs fuse mount and deleted those
files
right(not directly from the bricks)?
Can you restart nfs-ganesha server and see what happens ?
What type of volume are you using?
--
Jiffin
Post by Bernhard Dübi
OS: Ubuntu 16.04
Gluster: 3.8.13
Ganesha: 2.4.5
Let me know if you need more info
Best Regards
Bernhard
_______________________________________________
Gluster-users mailing list
http://lists.gluster.org/mailman/listinfo/gluster-users
Bernhard Dübi
2017-07-18 07:56:26 UTC
Reply
Permalink
Raw Message
Hi Jiffin,

thank you for the explanation

Kind Regards
Bernhard
Post by Jiffin Tony Thottan
Post by Bernhard Dübi
Hi,
both Gluster servers were rebooted and now the unlink directory is clean.
Following should have happened, If delete operation is performed gluster
keeps file in .unlink directory if it has open fd.
In this case since lazy umount is performed, ganesha server may still keep
the fd's open by that client so gluster keeps
the unlink directory even though it is removed from fuse mount.
--
Jiffin
Post by Bernhard Dübi
Best Regards
Bernhard
Post by Bernhard Dübi
Hi,
yes, I mounted the Gluster volume and deleted the files from the
volume not the brick
mount -t glusterfs hostname:volname /mnt
cd /mnt/some/directory
rm -rf *
restart of nfs-ganesha is planned for tomorrow. I'll keep you posted
BTW: nfs-ganesha is running on a separate server in standalone configuration
Best Regards
Bernhard
Post by Bernhard Dübi
Hello everybody,
I'm in a similar situation as described in
https://bugzilla.redhat.com/show_bug.cgi?id=1374166
The issue got fixed by https://review.gluster.org/#/c/14820 and is already
available in 3.8 branch
Post by Bernhard Dübi
I have a gluster volume exported through ganesha. we had some problems
on the gluster server and the NFS mount on the client was hanging.
I did a lazy umount of the NFS mount on the client, then went to the
Gluster server, mounted the Gluster volume and deleted a bunch of
files.
When I mounted the volume again on the client I noticed that the space
was not freed. Now I find them in $brick/.glusterfs/unlink
Here you have mounted the volume via glusterfs fuse mount and deleted those
files
right(not directly from the bricks)?
Can you restart nfs-ganesha server and see what happens ?
What type of volume are you using?
--
Jiffin
Post by Bernhard Dübi
OS: Ubuntu 16.04
Gluster: 3.8.13
Ganesha: 2.4.5
Let me know if you need more info
Best Regards
Bernhard
_______________________________________________
Gluster-users mailing list
http://lists.gluster.org/mailman/listinfo/gluster-users
Loading...