Allow "mds cache memory limit" tuning
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Ceph-FS Charm |
Fix Committed
|
Wishlist
|
Ethan Myers |
Bug Description
[Impact]
In the luminous release, ceph added support for modifying the metadata server cache memory limits [0].
On larger deployments, adjusting this setting upwards from the 1GB default is necessary to meet ceph fs client needs.
By default, ceph reserves 5% of the cache limit for the creation of new metadata. Make sure to account for this reserved amount when calculating suitable cache sizings.
[Example]
# Query the active configuration on the mds host
sudo ceph daemon mds.$HOSTNAME config show | grep memory_limit
# Add the following lines to ceph.conf persist a new 8GB cache limit
[mds]
mds cache memory limit = 8589934592
# Run the following command to apply the new run-time setting
sudo ceph daemon mds.$HOSTNAME config set
[Other Info]
The "mds cache memory limit" setting is configurable at run-time [1]. Apply through `ceph config` (Mimic+) or through `ceph tell` or `ceph daemon` (Luminous+). A service restart is not necessary to pick up the change.
[0] https:/
[1] https:/
Changed in charm-ceph-fs: | |
status: | New → Confirmed |
importance: | Undecided → Wishlist |
Changed in charm-ceph-fs: | |
assignee: | nobody → Ethan Myers (ethanmyers) |
Changed in charm-ceph-fs: | |
status: | Confirmed → In Progress |
Fix proposed to branch: master /review. opendev. org/c/openstack /charm- ceph-fs/ +/842555
Review: https:/