This change causes directory handles to always report an mtime of
"now." This is not an invalidate per se--it interacts with the
nfs implementation to produce that result when the implementation
updates its cached attributes. Hence, it can be modulated by timers
or other rules governing attribute caching at the upper layer.
Fixes: http://tracker.ceph.com/issues/40204
Signed-off-by: Matt Benjamin <mbenjamin@redhat.com>
two reasons
* GCC-9 features more checks. so let's use it!
* crimson targets the hardware + toolchain + kernel after 1+ years,
so it would be great if we can compile and test crimson using
the toolchain which is ubiquitous on most mainstream distos.
Signed-off-by: Kefu Chai <kchai@redhat.com>
this package is commented out using `# Crimson`. and it will be enabled
only if we are building "crimson" flavor builds.
this package depends on ceph-osd, because it'll basically be a drop-in
replacement of ceph-osd. and it shares all the dependencies of ceph-osd.
so let's just reuse ceph-osd package for now.
Signed-off-by: Kefu Chai <kchai@redhat.com>
because `install-deps.sh` is executed using `source`, we have to pass
these options using env variables. but before this change, `WITH_SEASTAR` is used directly,
while `FOR_MAKE_CHECK` is checked and translated to a local variable
`for_make_check`. which, in my opinion, has better readability.
so, in this change, `WITH_SEASTAR` is translated to `with_seastar`
variable in `install-deps.sh`.
Signed-off-by: Kefu Chai <kchai@redhat.com>
Originally, root_fh was reachable from RGWLibFS::fh_cache, but
this was problematic as it has infinite lifetime and so is not
present in RGWLibFS::fs_lru.
To fix the above, RGWLibFS::root_fh came to live unhooked from the
handle caches which works because nfs-ganesha/the application
always has an instance handle. However, this silently breaks
READDIR event invalidation.
Fixes: https://tracker.ceph.com/issues/40196
Signed-off-by: Matt Benjamin <mbenjamin@redhat.com>
* refs/pull/28363/head:
osd/ClassHandler: use std::variant for storing func
osd/ClassHandler: no need to add `struct` before struct type
osd/ClassHandler: use ceph::mutex instead of Mutex
osd/ClassHandler: remove unused function
osd/ClassHandler: s/string/std::string/ in header
Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Radoslaw Zarzynski <rzarzyns@redhat.com>
As a daemon, use the build-time specified alloc library
(tcmalloc/jemalloc) when linking the application.
Fixes: http://tracker.ceph.com/issues/40110
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
Instead of bare 'except', catch OSError and libcephfs exceptions.
Fixes: https://tracker.ceph.com/issues/40191
Signed-off-by: Varsha Rao <varao@redhat.com>
Break the line into multiple lines and add new variable wherever necessary, to
fix the line too long error.
Fixes: https://tracker.ceph.com/issues/40191
Signed-off-by: Varsha Rao <varao@redhat.com>
the list of buffers is protected by mutex m_lock. when dump_data() fails
and puts unwritten buffers back into the list, it needs to reaquire the
lock
Fixes: http://tracker.ceph.com/issues/40188
Signed-off-by: Casey Bodley <cbodley@redhat.com>
MDS purge queue didn't have a perf counter to record how many items
still left in journal. Even when MDS restarted, there was no any hint to
know how many inodes haven't been really deleted from disks.
Fixes: http://tracker.ceph.com/issues/40121
Signed-off-by: Zhi Zhang <zhangz.david@outlook.com>
With parallel clean_pg_upmaps feature on, the total time cost
of the performance test which now can utilize up to 8 threads for
parallel upmap validating decreased from:
maybe_remove_pg_upmaps (~10000 pg_upmap_items) latency:104s
to:
clean_pg_upmaps (~10000 pg_upmap_items) latency:7s
Note that by default the mon uses only 4 worker threads for
CPU intensive background work, you could further increase
the "mon_cpu_threads" option value if you decided the
time-consuming of clean_pg_upmaps still matters.
Fixes: http://tracker.ceph.com/issues/40104
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
There could definitely be some certain cases we could reliably
skip this kind of checking, but there is no easy way to separate
those out.
However, this is clearly the general way to do the massive pg
upmap clean-up job more efficiently and hence should make sense
in all cases.
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
- it's good to read.
- the updating pending_inc part should be made independent
since it is going to be racy while running in parallel.
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
We hit a couple more SELinux denials when running ceph on RHEL8. The
dac_read_search change is related to a kernel change where it checks
dac_read_search before dac_override, now.
Signed-off-by: Boris Ranto <branto@redhat.com>
