4

I am using MHDDFS to pool a bunch of large disks together into a single filesystem that I can easily share across a network.

For some reason every now and then the virtual filesystem seems to fall apart and I get a "cannot access datastore: Transport endpoint is not connected" whenever I perform an ls command in the directory containing the virtual filesystem.

As a result the filesystem becomes unusable accross the network. I have clients accessing it via SAMBA (windows) and NFS.

I have seen many users complain about this online but all the suggested solutions are unacceptable to me. Most people suggest unmounting and remounting the filesystem. A machine restart also works. I don't want to do this because then I have to SSH into the box every time this happens.

I'd like to know why this happens. The MHDDFS logfile is useless since it simply stops at the last file access entry and does not say what went wrong.

Any advice would be appreciated!

getack
  • 61
  • 8
  • Are there any other relevant seeming logs in syslog or dmesg? – Grant Mar 22 '15 at 12:42
  • Thank you for getting back to me! I will grok the logs asap and update my question with any relevant findings. I reread the MHDDFS manpage and saw that the default loglevel is not debug. So I will reenable it tonight to see if I can learn anything. Leaving logging on the standard loglevel gives me a <50GB (!!) logfile after 7 days though. – getack Mar 23 '15 at 13:16
  • at least while you are having issues, turn on the debug log. Make sure it is NOT stored inside the mhddfs mount. Also, check health of all drives with smartctl - many mhddfs issues i have had were triggered by small issues in the underlying drives. – Grant Mar 23 '15 at 15:44

2 Answers2

2

This appears to be a bug in version 0.1.39 of mhddfs on Ubuntu. A workaround is to downgrade to 0.1.38 from Precise.

Michael Hampton
  • 244,070
  • 43
  • 506
  • 972
user2296273
  • 31
  • 1
  • I am not sure why this is downvoted, but I will do this tonight. With my luck after enabling the debug logging MHDDFS have become rock solid with no crashes for almost a week! – getack Mar 30 '15 at 06:01
2

I think I've found the bug and have a patch out. https://github.com/ram-nat/mhddfs/commit/26d0f119eaa7e3ffaaf330bf29672e13471cb091

Details: http://nramkumar.org/tech/blog/2015/09/23/mhddfs-crash-with-ubuntu-14-04/

Ram
  • 21
  • 1