Red Hat Bugzilla – Bug 1296824
Need a deterministic way to account how much read/write would constitute a one read/write threshold
Last modified: 2018-02-06 12:43:16 EST
Description of problem:
Setting of cluster.read-freq-threshold is not honored. While cluster.read-freq-threshold is set to a value more than 1, files are promoted even when they are read once.
For the test, cluster.read-freq-threshold was set to 3, and a file from cold tier was accessed by fuse mount once. File got promoted.
[root@zod /]# gluster vol info
Volume Name: krk-vol
Volume ID: 1cb424ba-8f21-41f7-8d83-6c321d79a15c
Number of Bricks: 10
Hot Tier :
Hot Tier Type : Distributed-Replicate
Number of Bricks: 2 x 2 = 4
Cold Tier Type : Distribute
Number of Bricks: 6
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. create a distributed gluster volume
2. create few files
3. attach a hot tier
4. enable quota, turn performance.io-cache and performance.quick-read off
5. configure cluster.read-freq-threshold to a higher value
5. read files that were created earlier in step 2 (cat and vi were used for the test). Read only once.
Files that were read once were promoted
Files should be read at least 'n' number of times to be promoted, where n is the value set for cluster.read-freq-threshold
cluster.write-freq-threshold works as expected
sosreports are available here --> http://rhsqe-repo.lab.eng.blr.redhat.com/sosreports/1296824/
Thank you for your bug report.
We are no longer working on any improvements for Tier. This bug will be set to CLOSED WONTFIX to reflect this. Please reopen if the rfe is deemed critical.