Bug #373
closed
cfuse is degenerate with respect to ptmalloc
Added by Greg Farnum over 13 years ago.
Updated over 7 years ago.
Description
While running the test suite on cfuse I saw real memory usage climb up to 2GB. That doesn't quite seem correct. Determine if it's a memory leak or a heap fragmentation issue and try to fix.
- Target version changed from v0.23 to v0.22
While running the qa suite on flab, it's now up to 1.4GB. Tests run so far:
snaptest-xattrwb
snaptest-1
snaptest-2
iozone
In progress: snaptest-dir-rename
If we're lucky it's just one of these doing it, and I can identify the single test and run it through valgrind or something.
Oh, I misread that. Also completed: untar_snap_rm. In progress: kernel_untar_build
- Subject changed from Memory leak in cfuse to cfuse is degenerate with respect to ptmalloc
Well, I ran it through massif. And it looks like cfuse, as with the MDS and OSD, is just degenerate for ptmalloc. massif reported peak usage of ~400MB. I tried switching to cfuse and ran the iozone test and it ended at ~650MB virtual/545MB resident. In previous tests I've seen it top 1.6GB virtual memory.
I'll have to look into how we use the client to see if linking it with tcmalloc is a viable solution.
- Status changed from New to Resolved
Building cfuse with tcmalloc as of commit:23b1b52b80ed45c17ad3a67f2645381eac042455
- Project changed from Ceph to CephFS
- Category deleted (
11)
- Target version deleted (
v0.22)
Bulk updating project=ceph category=ceph-fuse issues to move to fs project so that we can remove the ceph-fuse category from the ceph project
Also available in: Atom
PDF