Hi,
We're still seeing this issue, we can see entries like the below on the servers connecting to gluster in the gluster client logs in case that helps:
[2023-07-06 22:59:47.029036 +0000] W [MSGID: 114031] [client-rpc-fops_v2.c:2561:client4_0_lookup_cbk] 0-gv1-client-7: remote operation failed. [{path=<gfid:55bca2d2-ae47-4
57c-82f2-bcea5947f558>}, {gfid=55bca2d2-ae47-457c-82f2-bcea5947f558}, {errno=2}, {error=No such file or directory}]
[2023-07-06 22:59:47.029136 +0000] W [MSGID: 114031] [client-rpc-fops_v2.c:2561:client4_0_lookup_cbk] 0-gv1-client-5: remote operation failed. [{path=<gfid:55bca2d2-ae47-4
57c-82f2-bcea5947f558>}, {gfid=55bca2d2-ae47-457c-82f2-bcea5947f558}, {errno=2}, {error=No such file or directory}]
[2023-07-06 22:59:47.038752 +0000] E [MSGID: 133021] [shard.c:3822:shard_delete_shards] 0-gv1-shard: Failed to clean up shards of gfid 55bca2d2-ae47-457c-82f2-bcea5947f558
[No such file or directory]
Thanks,
| |
| Liam Smith |
| Linux Systems Support Engineer, Scholar |
|
|
|
|
|
|
|
From: Strahil Nikolov <hunter86_bg@xxxxxxxxx>
Sent: 05 July 2023 11:59
To: Liam Smith <liam.smith@xxxxx>; gluster-users@xxxxxxxxxxx <gluster-users@xxxxxxxxxxx>; Gluster Devel <gluster-devel@xxxxxxxxxxx>
Subject: Re: [Gluster-users] remove_me files building up
|
CAUTION: This e-mail originates from outside of Ekco. Do not click
links or attachments unless you recognise the sender.
|
Adding Gluster Devel list.
Best Regards,
Strahil Nikolov
On Wednesday, July 5, 2023, 12:41 PM, Liam Smith <liam.smith@xxxxx> wrote:
The contents of this email message and any attachments are intended solely for the addressee(s) and may contain confidential and/or privileged information and may be legally protected from disclosure.
Hi Strahil,
This is the output from the commands:
root@uk3-prod-gfs-arb-01:~# du -h -x -d 1 /data/glusterfs/gv1/brick1/brick
2.2G /data/glusterfs/gv1/brick1/brick/.glusterfs
24M /data/glusterfs/gv1/brick1/brick/scalelite-recordings
16K /data/glusterfs/gv1/brick1/brick/mytute
18M /data/glusterfs/gv1/brick1/brick/.shard
0 /data/glusterfs/gv1/brick1/brick/.glusterfs-anonymous-inode-d3d1fdec-7df9-4f71-b9fc-660d12c2a046
2.3G /data/glusterfs/gv1/brick1/brick
root@uk3-prod-gfs-arb-01:~# du -h -x -d 1 /data/glusterfs/gv1/brick3/brick
11G /data/glusterfs/gv1/brick3/brick/.glusterfs
15M /data/glusterfs/gv1/brick3/brick/scalelite-recordings
460K /data/glusterfs/gv1/brick3/brick/mytute
151M /data/glusterfs/gv1/brick3/brick/.shard
0 /data/glusterfs/gv1/brick3/brick/.glusterfs-anonymous-inode-d3d1fdec-7df9-4f71-b9fc-660d12c2a046
11G /data/glusterfs/gv1/brick3/brick
root@uk3-prod-gfs-arb-01:~# du -h -x -d 1 /data/glusterfs/gv1/brick2/brick
12G /data/glusterfs/gv1/brick2/brick/.glusterfs
110M /data/glusterfs/gv1/brick2/brick/scalelite-recordings
3.1M /data/glusterfs/gv1/brick2/brick/mytute
169M /data/glusterfs/gv1/brick2/brick/.shard
0 /data/glusterfs/gv1/brick2/brick/.glusterfs-anonymous-inode-d3d1fdec-7df9-4f71-b9fc-660d12c2a046
12G /data/glusterfs/gv1/brick2/brick
Also, this is the du -sh output for the specific directory that appears to be taking up space on brick 3:
root@uk3-prod-gfs-arb-01:~# du -sh /data/glusterfs/gv1/brick3/brick/.shard/.remove_me/
10G /data/glusterfs/gv1/brick3/brick/.shard/.remove_me/
The gluster package version on all the servers is 11.0, and I believe they were upgraded from 7.2>8.6>11.0; the op.version was only changed after the 11.0 upgrade.
The archival job deletes the files, files shouldn't be overwritten at any point as it's new, unique files being generated everyday.
Thanks,
| |
| Liam Smith |
| Linux Systems Support Engineer, Scholar |
|
|
|
|
|
|
|
From: Strahil Nikolov <hunter86_bg@xxxxxxxxx>
Sent: 04 July 2023 17:47
To: Liam Smith <liam.smith@xxxxx>; gluster-users@xxxxxxxxxxx <gluster-users@xxxxxxxxxxx>
Subject: Re: [Gluster-users] remove_me files building up
|
CAUTION: This e-mail originates from outside of Ekco. Do not click
links or attachments unless you recognise the sender.
|
Thanks for the clarification.
That behaviour is quite weird as arbiter bricks should hold only metadata.
What does the following show on host uk3-prod-gfs-arb-01:
du -h -x -d 1 /data/glusterfs/gv1/brick1/brick
du -h -x -d 1 /data/glusterfs/gv1/brick3/brick
du -h -x -d 1 /data/glusterfs/gv1/brick2/brick
If indeed the shards are taking space - that is a really strange situation.
From which version did you upgrade and which one is now ? I assume all gluster TSP members
(the servers) have the same version, but it’s nice to double check.
Does the archival job actually deletes the original files after being processed or the workload keeps overriding the existing files ?
Best Regards,
Strahil Nikolov
On Tuesday, July 4, 2023, 6:50 PM, Liam Smith <liam.smith@xxxxx> wrote:
The contents of this email message and any attachments are intended solely for the addressee(s) and may contain confidential and/or privileged information and may be legally protected from disclosure.
Hi Strahil,
We're using gluster to act as a share for an application to temporarily process and store files, before they're then archived off over night.
The issue we're seeing isn't with the inodes running out of space, but the actual disk space on the arb server running low.
This is the df -h output for the bricks on the arb server:
/dev/sdd1 15G 12G 3.3G 79% /data/glusterfs/gv1/brick3
/dev/sdc1 15G 2.8G 13G 19% /data/glusterfs/gv1/brick1
/dev/sde1 15G 14G 1.6G 90% /data/glusterfs/gv1/brick2
And this is the df -hi output for the bricks on the arb server:
/dev/sdd1 7.5M 2.7M 4.9M 35% /data/glusterfs/gv1/brick3
/dev/sdc1 7.5M 643K 6.9M 9% /data/glusterfs/gv1/brick1
/dev/sde1 6.1M 3.0M 3.1M 49% /data/glusterfs/gv1/brick2
So the inode usage appears to be fine, but we're seeing that the actual disk usage keeps increasing on the bricks despite it being the arbiter.
The actual issue appears to be that files under /data/glusterfs/gv1/brick3/brick/.shard/.remove_me/ and
/data/glusterfs/gv1/brick2/brick/.shard/.remove_me/ are being retained, even when the original files are
deleted from the data nodes.
For reference, I've attached disk usage graphs for brick 3 over the past two weeks; one is a graph from a data node, the other from the arb.
As you can see, the disk usage of the data node builds throughout the day, but then an archival job clears space down. However, on the arb, we see the disk space increasing in the same sort of trend, but it's never cleared down like the data node.
Hopefully this clarifies the issue, we're a bit confused as to why this is occurring and whether this is actually intended behaviour or potentially a bug, so any advice is greatly appreciated.
Thanks,
| |
| Liam Smith |
| Linux Systems Support Engineer, Scholar |
|
|
|
|
The contents of this email message and any attachments are intended solely for the addressee(s) and may contain confidential and/or privileged information and may be legally protected from disclosure.
|
|
From: Strahil Nikolov <hunter86_bg@xxxxxxxxx>
Sent: 04 July 2023 15:51
To: Liam Smith <liam.smith@xxxxx>; gluster-users@xxxxxxxxxxx <gluster-users@xxxxxxxxxxx>
Subject: Re: [Gluster-users] remove_me files building up
|
CAUTION: This e-mail originates from outside of Ekco. Do not click
links or attachments unless you recognise the sender.
|
Hi Liam,
I saw that your XFS uses ‘ imaxpct=25’ which for an arbiter brick is a little bit low.
If you have free space on the bricks, increase the maxpct to a bigger value, like:
xfs_growfs -m 80 /path/to/brick
That will set 80% of the Filesystem for inodes, which you can verify with df -i /brick/path (compare before and after). This way you won’t run out of inodes in the future.
Of course, always test that on non Prod first.
Are you using the volume for VM disk storage domain ? What is your main workload ?
Best Regards,
Strahil Nikolov
On Tuesday, July 4, 2023, 2:12 PM, Liam Smith <liam.smith@xxxxx> wrote:
Hi,
Thanks for your response, please find the xfs_info for each brick on the arbiter below:
root@uk3-prod-gfs-arb-01:~# xfs_info /data/glusterfs/gv1/brick1
meta-data="" isize=512 agcount=31, agsize=131007 blks
= sectsz=512 attr=2, projid32bit=1
= crc=1 finobt=1, sparse=1, rmapbt=0
= reflink=1
data = bsize=4096 blocks=3931899, imaxpct=25
= sunit=0 swidth=0 blks
naming =version 2 bsize=4096 ascii-ci=0, ftype=1
log =internal log bsize=4096 blocks=2560, version=2
= sectsz=512 sunit=0 blks, lazy-count=1
realtime =none extsz=4096 blocks=0, rtextents=0
root@uk3-prod-gfs-arb-01:~# xfs_info /data/glusterfs/gv1/brick2
meta-data="" isize=512 agcount=13, agsize=327616 blks
= sectsz=512 attr=2, projid32bit=1
= crc=1 finobt=1, sparse=1, rmapbt=0
= reflink=1
data = bsize=4096 blocks=3931899, imaxpct=25
= sunit=0 swidth=0 blks
naming =version 2 bsize=4096 ascii-ci=0, ftype=1
log =internal log bsize=4096 blocks=2560, version=2
= sectsz=512 sunit=0 blks, lazy-count=1
realtime =none extsz=4096 blocks=0, rtextents=0
root@uk3-prod-gfs-arb-01:~# xfs_info /data/glusterfs/gv1/brick3
meta-data="" isize=512 agcount=13, agsize=327616 blks
= sectsz=512 attr=2, projid32bit=1
= crc=1 finobt=1, sparse=1, rmapbt=0
= reflink=1
data = bsize=4096 blocks=3931899, imaxpct=25
= sunit=0 swidth=0 blks
naming =version 2 bsize=4096 ascii-ci=0, ftype=1
log =internal log bsize=4096 blocks=2560, version=2
= sectsz=512 sunit=0 blks, lazy-count=1
realtime =none extsz=4096 blocks=0, rtextents=0
I've also copied below some df output from the arb server:
root@uk3-prod-gfs-arb-01:~# df -hi
Filesystem Inodes IUsed IFree IUse% Mounted on
udev 992K 473 991K 1% /dev
tmpfs 995K 788 994K 1% /run
/dev/sda1 768K 105K 664K 14% /
tmpfs 995K 3 995K 1% /dev/shm
tmpfs 995K 4 995K 1% /run/lock
tmpfs 995K 18 995K 1% /sys/fs/cgroup
/dev/sdb1 128K 113 128K 1% /var/lib/glusterd
/dev/sdd1 7.5M 2.6M 5.0M 35% /data/glusterfs/gv1/brick3
/dev/sdc1 7.5M 600K 7.0M 8% /data/glusterfs/gv1/brick1
/dev/sde1 6.4M 2.9M 3.5M 46% /data/glusterfs/gv1/brick2
uk1-prod-gfs-01:/gv1 150M 6.5M 144M 5% /mnt/gfs
tmpfs 995K 21 995K 1% /run/user/1004
root@uk3-prod-gfs-arb-01:~# df -h
Filesystem Size Used Avail Use% Mounted on
udev 3.9G 0 3.9G 0% /dev
tmpfs 796M 916K 795M 1% /run
/dev/sda1 12G 3.9G 7.3G 35% /
tmpfs 3.9G 8.0K 3.9G 1% /dev/shm
tmpfs 5.0M 0 5.0M 0% /run/lock
tmpfs 3.9G 0 3.9G 0% /sys/fs/cgroup
/dev/sdb1 2.0G 456K 1.9G 1% /var/lib/glusterd
/dev/sdd1 15G 12G 3.5G 78% /data/glusterfs/gv1/brick3
/dev/sdc1 15G 2.6G 13G 18% /data/glusterfs/gv1/brick1
/dev/sde1 15G 14G 1.8G 89% /data/glusterfs/gv1/brick2
uk1-prod-gfs-01:/gv1 300G 139G 162G 47% /mnt/gfs
tmpfs 796M 0 796M 0% /run/user/1004
Something I forgot to mention in my initial message is that the opversion was upgraded from 70200 to 100000, which seems as though it could have been a trigger for the issue as well.
Thanks,
| |
|
Liam Smith |
|
Linux Systems Support Engineer, Scholar |
|
|
|
|
|
|
|
From: Strahil Nikolov <hunter86_bg@xxxxxxxxx>
Sent: 03 July 2023 18:28
To: Liam Smith <liam.smith@xxxxx>; gluster-users@xxxxxxxxxxx <gluster-users@xxxxxxxxxxx>
Subject: Re: [Gluster-users] remove_me files building up
|
CAUTION: This e-mail originates from outside of Ekco. Do not click
links or attachments unless you recognise the sender.
|
Hi,
you mentioned that the arbiter bricks run out of inodes.
Are you using XFS ?
Can you provide the xfs_info of each brick ?
Best Regards,
Strahil Nikolov
The contents of this email message and any attachments are intended solely for the addressee(s) and may contain confidential and/or privileged information and may be legally protected from disclosure.
On Sat, Jul 1, 2023 at 19:41, Liam Smith
<liam.smith@xxxxx> wrote:
Hi,
We're running a cluster with two data nodes and one arbiter, and have sharding enabled.
We had an issue a while back where one of the server's crashed, we got the server back up and running and ensured that all healing entries cleared, and also increased the server spec (CPU/Mem) as this seemed to be the potential cause.
Since then however, we've seen some strange behaviour, whereby a lot of 'remove_me' files are building up under `/data/glusterfs/gv1/brick2/brick/.shard/.remove_me/` and `/data/glusterfs/gv1/brick3/brick/.shard/.remove_me/`. This
is causing the arbiter to run out of space on brick2 and brick3, as the remove_me files are constantly increasing.
brick1 appears to be fine, the disk usage increases throughout the day and drops down in line with the trend of the brick on the data nodes. We see the disk usage increase and drop throughout the day on the data nodes for brick2 and brick3 as well, but while
the arbiter follows the same trend of the disk usage increasing, it doesn't drop at any point.
This is the output of some gluster commands, occasional heal entries come and go:
root@uk3-prod-gfs-arb-01:~# gluster volume info gv1
Volume Name: gv1
Type: Distributed-Replicate
Volume ID: d3d1fdec-7df9-4f71-b9fc-660d12c2a046
Status: Started
Snapshot Count: 0
Number of Bricks: 3 x (2 + 1) = 9
Transport-type: tcp
Bricks:
Brick1: uk1-prod-gfs-01:/data/glusterfs/gv1/brick1/brick
Brick2: uk2-prod-gfs-01:/data/glusterfs/gv1/brick1/brick
Brick3: uk3-prod-gfs-arb-01:/data/glusterfs/gv1/brick1/brick (arbiter)
Brick4: uk1-prod-gfs-01:/data/glusterfs/gv1/brick3/brick
Brick5: uk2-prod-gfs-01:/data/glusterfs/gv1/brick3/brick
Brick6: uk3-prod-gfs-arb-01:/data/glusterfs/gv1/brick3/brick (arbiter)
Brick7: uk1-prod-gfs-01:/data/glusterfs/gv1/brick2/brick
Brick8: uk2-prod-gfs-01:/data/glusterfs/gv1/brick2/brick
Brick9: uk3-prod-gfs-arb-01:/data/glusterfs/gv1/brick2/brick (arbiter)
Options Reconfigured:
cluster.entry-self-heal: on
cluster.metadata-self-heal: on
cluster.data-self-heal: on
performance.client-io-threads: off
storage.fips-mode-rchecksum: on
transport.address-family: inet
cluster.lookup-optimize: off
performance.readdir-ahead: off
cluster.readdir-optimize: off
cluster.self-heal-daemon: enable
features.shard: enable
features.shard-block-size: 512MB
cluster.min-free-disk: 10%
cluster.use-anonymous-inode: yes
root@uk3-prod-gfs-arb-01:~# gluster peer status
Number of Peers: 2
Hostname: uk2-prod-gfs-01
Uuid: 2fdfa4a2-195d-4cc5-937c-f48466e76149
State: Peer in Cluster (Connected)
Hostname: uk1-prod-gfs-01
Uuid: 43ec93d1-ad83-4103-aea3-80ded0903d88
State: Peer in Cluster (Connected)
root@uk3-prod-gfs-arb-01:~# gluster volume heal gv1 info
Brick uk1-prod-gfs-01:/data/glusterfs/gv1/brick1/brick
<gfid:5b57e1f6-3e3d-4334-a0db-b2560adae6d1>
Status: Connected
Number of entries: 1
Brick uk2-prod-gfs-01:/data/glusterfs/gv1/brick1/brick
Status: Connected
Number of entries: 0
Brick uk3-prod-gfs-arb-01:/data/glusterfs/gv1/brick1/brick
Status: Connected
Number of entries: 0
Brick uk1-prod-gfs-01:/data/glusterfs/gv1/brick3/brick
Status: Connected
Number of entries: 0
Brick uk2-prod-gfs-01:/data/glusterfs/gv1/brick3/brick
Status: Connected
Number of entries: 0
Brick uk3-prod-gfs-arb-01:/data/glusterfs/gv1/brick3/brick
Status: Connected
Number of entries: 0
Brick uk1-prod-gfs-01:/data/glusterfs/gv1/brick2/brick
Status: Connected
Number of entries: 0
Brick uk2-prod-gfs-01:/data/glusterfs/gv1/brick2/brick
<gfid:6ba9c472-9232-4b45-b12f-a1232d6f4627>
/.shard/.remove_me
<gfid:0f042518-248d-426a-93f4-cfaa92b6ef3e>
Status: Connected
Number of entries: 3
Brick uk3-prod-gfs-arb-01:/data/glusterfs/gv1/brick2/brick
<gfid:6ba9c472-9232-4b45-b12f-a1232d6f4627>
/.shard/.remove_me
<gfid:0f042518-248d-426a-93f4-cfaa92b6ef3e>
Status: Connected
Number of entries: 3
root@uk3-prod-gfs-arb-01:~# gluster volume get all cluster.op-version
Option Value
------ -----
cluster.op-version 100000
We're not sure if this is a potential bug or if something's corrupted that we don't have visibility of, so any pointers/suggestions about how to approach this would be appreciated.
Thanks,
Liam
The contents of this email message and any attachments are intended solely for the addressee(s) and may contain confidential and/or privileged information and may be legally protected from disclosure.
________
Community Meeting Calendar:
Schedule -
Every 2nd and 4th Tuesday at 14:30 IST / 09:00 UTC
Bridge:
https://meet.google.com/cpu-eiue-hvk
Gluster-users mailing list
Gluster-users@xxxxxxxxxxx
https://lists.gluster.org/mailman/listinfo/gluster-users
|
