[Gluster-users] Files exist, but sometimes are not seen by the clients: "No such file or directory"

Carlo Rodrigues carlo.rodrigues at pt.clara.net
Wed Jan 3 19:54:55 UTC 2024


Hello all,

We're having problems with files that suddenly stop being seen on the fuse clients.
I couldn't yet find a way to reproduce this. It happens every once in a while.

Sometimes you try to ls some file and it can't be found.
When you run ls on the parent directory, it is shown on the output, and, after that, you can access it.
I'm mentioning ls, but the problem also manifests when trying to access the file in any other way, not only when listing it with the ls command.

Example: (The file names and location were edited for brevity)

1. ls file returns "No such file or directory"

$ ls -l /home/RESEARCH/user.x/some_path/idtracker.log
ls: cannot access /home/RESEARCH/user.x/some_path/idtracker.log: No such file or directory

2. ls parent directory. The file is shown in the results.

$ ls -l /home/RESEARCH/user.x/some_path/
total 334500
-rw-r--r-- 1 user.x group_x 348521772 Jan 19 15:26 file_17.avi
-rw-r--r-- 1 user.x group_x    978252 Jan 19 15:26 file_17.csv
-rw-r--r-- 1 user.x group_x      1185 Jun  5 10:05 idtracker.log
drwxr-xr-x 2 user.x group_x      4096 Jun  2 21:17 segm

3. try to ls file again. This time, it is found.

$ ls -l /home/RESEARCH/user.x/some_path/idtracker.log
-rw-r--r-- 1 user.x group_x 1185 Jun  5 10:05 /home/RESEARCH/user.x/some_path/idtracker.log


The glusterfs filesystem is mounted like this
storage:/tank on /nfs/tank type fuse.glusterfs (rw,relatime,user_id=0,group_id=0,default_permissions,allow_other,max_read=131072)

At the moment, cluster.readdir-optimize was on, cluster.lookup-optimize was on and cluster.lookup-unhashed was off.

Now I've changed those 3 options to check if the problem still appears.

Current configuration:

$ sudo gluster volume info tank

Volume Name: tank
Type: Distribute
Volume ID: 2c3ca623-c1df-4b5f-b20e-5222d189f834
Status: Started
Snapshot Count: 0
Number of Bricks: 24
Transport-type: tcp
Bricks:
Brick1: swp-gluster-05:/tank/volume1/brick
Brick2: swp-gluster-06:/tank/volume1/brick
Brick3: swp-gluster-07:/tank/volume1/brick
Brick4: swp-gluster-05:/tank/volume2/brick
Brick5: swp-gluster-06:/tank/volume2/brick
Brick6: swp-gluster-07:/tank/volume2/brick
Brick7: swp-gluster-05:/tank/volume3/brick
Brick8: swp-gluster-06:/tank/volume3/brick
Brick9: swp-gluster-07:/tank/volume3/brick
Brick10: swp-gluster-05:/tank/volume4/brick
Brick11: swp-gluster-06:/tank/volume4/brick
Brick12: swp-gluster-07:/tank/volume4/brick
Brick13: swp-gluster-01:/tank/volume1/brick
Brick14: swp-gluster-02:/tank/volume1/brick
Brick15: swp-gluster-03:/tank/volume1/brick
Brick16: swp-gluster-04:/tank/volume1/brick
Brick17: swp-gluster-01:/tank/volume2/brick
Brick18: swp-gluster-02:/tank/volume2/brick
Brick19: swp-gluster-03:/tank/volume2/brick
Brick20: swp-gluster-04:/tank/volume2/brick
Brick21: swp-gluster-01:/tank/volume3/brick
Brick22: swp-gluster-02:/tank/volume3/brick
Brick23: swp-gluster-03:/tank/volume3/brick
Brick24: swp-gluster-04:/tank/volume3/brick
Options Reconfigured:
cluster.lookup-unhashed: on
cluster.lookup-optimize: off
performance.aggregate-size: 1MB
performance.read-ahead-page-count: 16
performance.nl-cache-timeout: 600
performance.nl-cache: on
network.inode-lru-limit: 200000
performance.md-cache-timeout: 600
performance.cache-invalidation: on
performance.cache-samba-metadata: on
features.cache-invalidation-timeout: 600
features.cache-invalidation: on
performance.write-behind: off
performance.cache-size: 128MB
storage.fips-mode-rchecksum: on
transport.address-family: inet
nfs.disable: on
features.inode-quota: on
features.quota: on
server.event-threads: 32
client.event-threads: 16
cluster.readdir-optimize: off
performance.io-thread-count: 64
performance.readdir-ahead: on
performance.client-io-threads: on
performance.parallel-readdir: disable
performance.read-ahead: on
performance.stat-prefetch: on
performance.open-behind: off
features.quota-deem-statfs: on

Has anyone gone through this before?
Thanks in advance.

Carlo Rodrigues
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20240103/f1a488ce/attachment.html>


More information about the Gluster-users mailing list