posix_fallocate deadlocks on ntfs if more than three nodes are started #15910

1. 
   MarcoFalke commented at 3:46 pm on April 27, 2019: member

   Starting four nodes will make them deadlock on the posix_fallocate call (and thus respond with Loading block index... (-28) on the RPC forever)

   Steps to reproduce:

   • Create NTFS partition on hdd
   • Run the test runner with TMP=/path/to/the/ntfs/partition ./test/functional/test_runner.py -j 9

   posix_fallocate falls back to fallocate on ntfs according to http://man7.org/linux/man-pages/man3/posix_fallocate.3.html#NOTES, not sure if that is relevant

2. MarcoFalke added the label Block storage on Apr 27, 2019
3. MarcoFalke added the label good first issue on Apr 27, 2019
4. 
   hebasto commented at 3:28 pm on May 2, 2019: member

   @MarcoFalke What are the driver (kernel’s ntfs or ntfs-3g) and mount options?
5. 
   MarcoFalke commented at 6:15 pm on May 2, 2019: member

   Not sure how to get them I use files 3.32.0-stable to mount it.

   0$ mount -l | grep 2CC35B3543E3A1E4
   1/dev/sda4 on /run/media/marco/2CC35B3543E3A1E4 type fuseblk (rw,nosuid,nodev,relatime,user_id=0,group_id=0,default_permissions,allow_other,blksize=4096,uhelper=udisks2)
   

6. 
   hebasto commented at 8:52 pm on May 2, 2019: member

   @MarcoFalke I cannot reproduce this issue on Linux Mint 19.1. Maybe, it is dependent on distro? Assuming you are running Fedora ;) My machine has only 4 cores. Can this affect the test_runner.py behavior?

   Is it still an issue after merging PR #15650?

7. MarcoFalke removed the label good first issue on Oct 17, 2019
8. 
   MarcoFalke commented at 7:58 pm on May 8, 2020: member

   Looks like leveldb has issues with ntfs: E.g. #10787
9. MarcoFalke closed this on Mar 8, 2021

   ---

10. DrahtBot locked this on Aug 18, 2022

Contributors
MarcoFalke hebasto

Labels
Block storage