macos-2.0.0-rc6 Memory Leak?
Posted: Fri Apr 30, 2021 11:16 am
I have a raidz3 pool (90TB ish) and seem to have a memory leak. When accessing files on either the native ZFS file system or on zvols formatted w/ HFS+ the system will eventually reboot after exhausting all memory. A vm_stat in a while loop shows the wired down pages increasing more or less monotonically and then eventually swapins/outs starting to happen and then an eventual panic. If I run the system but with the pool exported I never run into any issues. So it seems to be related to the zfs subsystem/extension.
I can reproduce it by doing a read loop of copying files from the filesystem.
For instance, on a ZVOL formatted with HFS+ filled with media files, the following
will trigger the eventual crash. I did all I could to quiesce all other activity and looking at lsof saw no other activity on the volume.
Thoughts on how I can help track down where the memory leak could be coming from?
If it helps, my pool doesn't have all the latest features enabled. In particular, a zpool upgrade informs me the following aren't enabled:
Thanks!
-david
I can reproduce it by doing a read loop of copying files from the filesystem.
For instance, on a ZVOL formatted with HFS+ filled with media files, the following
- Code: Select all
find . -type f -exec cat {} > /dev/null \;
will trigger the eventual crash. I did all I could to quiesce all other activity and looking at lsof saw no other activity on the volume.
Thoughts on how I can help track down where the memory leak could be coming from?
If it helps, my pool doesn't have all the latest features enabled. In particular, a zpool upgrade informs me the following aren't enabled:
userobj_accounting
project_quota
redaction_bookmarks
redacted_datasets
bookmark_written
log_spacemap
livelist
device_rebuild
zstd_compress
draid
Thanks!
-david