If the score did not improve, we should clear out any
pending pg-upmap-primary mappings so they don't execute
in situations where the same incremental is used to balance
multiple pools (i.e. in the balancer mgr module).
Signed-off-by: Laura Flores <lflores@ibm.com>
Followup to 8a5553597c.
Streamlines some of the logic so pg upmap activity is properly
initalized, and updated in offline mode as well as online.
Signed-off-by: Laura Flores <lflores@ibm.com>
Read balancing may now be managed automatically via the balancer
manager module. Users may choose between two new modes: ``upmap-read``, which
offers upmap and read optimization simultaneously, or ``read``, which may be used
to only optimize reads. Existing balancer commands have also been added to
contain more information about read balancing.
Run the following commands to test the new automatic behavior:
`ceph balancer on` (on by default)
`ceph balancer mode <read|upmap-read>`
`ceph balancer status`
Run the following commands to test the new supervised behavior:
`ceph balancer off`
`ceph balancer mode <read|upmap-read>`
`ceph balancer eval` | `ceph balancer eval <pool-name>`
`ceph balancer eval-verbose` | `ceph balancer eval-verbose <pool-name>`
`ceph balancer optimize <plan-name>`
`ceph balancer show <plan-name>`
`ceph balancer eval <plan-name>`
`ceph balancer execute <plan-name>`
In the balancer module, there is also a new "self_test" function which tests
the module's basic functionality. This test can be triggered with the following
commands:
`ceph mgr module enable selftest`
`ceph mgr self-test module balancer`
Related Trello: https://trello.com/c/sWoKctzL/859-add-read-balancer-support-inside-the-balancer-module
Signed-off-by: Laura Flores <lflores@ibm.com>
Changes how the upmap balancer compares min_mon_release
to account for release names eventually wrapping around the alphabet.
Signed-off-by: Laura Flores <lflores@ibm.com>
crimson/osd: drop a foreign-copy to shard-0 for every pg operation
Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Chunmei Liu <chunmei.liu@intel.com>
Reviewed-by: Matan Breizman <mbreizma@redhat.com>
Mapping rbd images to nbd devices using ioctl interface is not
robust. It was discovered that the device size or the md5 checksum
of the nbd device was incorrect immediately after mapping using
ioctl method. When using the nbd netlink interface to map RBD images
the issue was not encountered. Switch to using nbd netlink interface
for mapping.
Fixes: https://tracker.ceph.com/issues/64063
Signed-off-by: Ramana Raja <rraja@redhat.com>
Pull the `io_context` and threads out of `AsioFrontend`, pass in a
reference to `io_context_pool` so it can be shut down at `AsioFrontend::join`
Signed-off-by: Adam C. Emerson <aemerson@redhat.com>
fix: resolve inconsistent judgment of osd_pg_stat_report_interval_max
Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Matan Breizman <Matan.Brz@gmail.com>
`RadosDriver` needs it, and since SAL generally uses `optional_yield`,
other stores are going to need it to implement that sensibly.
Signed-off-by: Adam C. Emerson <aemerson@redhat.com>
Add overloads of the constructor and `start` that take an init
function, for setting up the thread before dropping into
`io_context::run`.
Signed-off-by: Adam C. Emerson <aemerson@redhat.com>
Just because this is what Ceph's config uses and it saves a narrowing
conversion. If we want to set a max value on the thread count, we
should do it in config.
Signed-off-by: Adam C. Emerson <aemerson@redhat.com>
Followup to commit 265f727481.
Fixes this error mesage under GitHub Insights (https://github.com/ceph/ceph/network/updates):
```
The property '#/updates/1/commit-message/prefix' was not of a maximum string length of 15
```
Signed-off-by: Laura Flores <lflores@ibm.com>
When using rbd CLI to map the images to NBD devices via netlink,
any errors that arose during image resizing in netlink_resize()
were not logged. Switching the error logging from using cerr to
derr helps log the errors from netlink_resize().
Signed-off-by: Ramana Raja <rraja@redhat.com>
Include device identifier or cookie in the message sent to the kernel
to resize images mapped to NBD devices using netlink. Otherwise,
netlink_resize() fails and the size of the device isn't updated.
Fixes: https://tracker.ceph.com/issues/64139
Signed-off-by: Ramana Raja <rraja@redhat.com>
with arrow's default BUILD_WARNING_LEVEL, -Werror is added to cflags and
debug builds fail on a warning about _FORTIFY_SOURCE
Fixes: https://tracker.ceph.com/issues/63130
Signed-off-by: Casey Bodley <cbodley@redhat.com>
