You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
@traylenator@vaneldik As reported by CERN batch computing admins, attempting to mount the same repository multiple times simultaneously can result in multiple cvmfs processes and a stuck mount helper process:
The issue seems to be that CheckConcurrentMount checks for the repository socket, which is only created later. When using fuse2, the subsequent mount helpers would later fail with:
fuse: mountpoint is not empty
fuse: if you are sure this is safe, use the 'nonempty' mount option
failed to create Fuse channel
(monitor) monitor stopped [09-13-2023 11:36:28 CEST]
but fuse3 apparently changed this behavior and lets the other mount attempts succeed as well, so this problem only becomes apparent when cvmfs-fuse3 is installed. In any case, this can be properly handled by using a lockfile in CheckConcurrentMount.
EDIT: the connection to fuse3 is not so clear, this has been observed with only fuse2 as well
The text was updated successfully, but these errors were encountered:
vvolkl
changed the title
Multiple cvmfs processes after concurrent mounts (when using cvmfs-fuse3)
Multiple cvmfs processes after concurrent mounts
Nov 6, 2023
@traylenator @vaneldik As reported by CERN batch computing admins, attempting to mount the same repository multiple times simultaneously can result in multiple cvmfs processes and a stuck mount helper process:
The issue seems to be that
CheckConcurrentMount
checks for the repository socket, which is only created later. When using fuse2, the subsequent mount helpers would later fail with:but fuse3 apparently changed this behavior and lets the other mount attempts succeed as well, so this problem only becomes apparent when
cvmfs-fuse3
is installed. In any case, this can be properly handled by using a lockfile in CheckConcurrentMount.EDIT: the connection to fuse3 is not so clear, this has been observed with only fuse2 as well
The text was updated successfully, but these errors were encountered: