+++ This bug was initially created as a clone of Bug #1282076 +++ Description of problem: Make cache mode the default for tiering Version-Release number of selected component (if applicable): How reproducible: always Steps to Reproduce: 1. 2. 3. Actual results: test mode is default Expected results: cache mode is default Additional info: test mode is for internal QE testing and should not be used by customers. Test mode forces all unused data to be destaged from the hot tier when no IO is received, rendering the SSD unused / wasted. If the demote cycle time is made large as a workaround, data will stay on the hot tier but will not react to changing conditions, which means unused data may fill the hot tier resource. A long cycle times also means the system will stay static even when close to full, which will put the system into a degraded mode. The cache mode is supposed to keep data on the hot tier but also demote data quickly when the hot tier is close to full. The cycle time can be set to a desired reaction time to swap out lesser used data and bring in more frequently accessed data without flushing the entire SSD. --- Additional comment from Vijay Bellur on 2015-11-14 14:57:34 EST --- REVIEW: http://review.gluster.org/12581 (cluster/tier make cache mode default for tiered volumes) posted (#1) for review on master by Dan Lambright (dlambrig) --- Additional comment from Vijay Bellur on 2015-11-17 07:18:33 EST --- COMMIT: http://review.gluster.org/12581 committed in master by Dan Lambright (dlambrig) ------ commit dcd1ff344d242f64f3a5c579df97a050736e6633 Author: Dan Lambright <dlambrig> Date: Sat Nov 14 14:35:26 2015 -0500 cluster/tier make cache mode default for tiered volumes The default mode for tiered volumes must be cache. The current test mode was for engineering and should ordinarily not be used by customers. Change-Id: I20583f54a9269ce75daade645be18ab8575b0b9b BUG: 1282076 Signed-off-by: Dan Lambright <dlambrig> Reviewed-on: http://review.gluster.org/12581 Tested-by: Gluster Build System <jenkins.com> Reviewed-by: mohammed rafi kc <rkavunga>
With 3.7.5-7 build. Cache mode is the default now. vol info shows as below: [root@zod ~]# gluster v info metadata Volume Name: metadata Type: Tier Volume ID: 23dcc876-5343-4ab9-b320-844e30e4a67c Status: Started Number of Bricks: 16 Transport-type: tcp Hot Tier : Hot Tier Type : Distributed-Replicate Number of Bricks: 2 x 2 = 4 Brick1: yarrow:/rhs/brick6/metadata_hot Brick2: zod:/rhs/brick6/metadata_hot Brick3: yarrow:/rhs/brick7/metadata_hot Brick4: zod:/rhs/brick7/metadata_hot Cold Tier: Cold Tier Type : Distributed-Disperse Number of Bricks: 2 x (4 + 2) = 12 Brick5: zod:/rhs/brick1/metadata Brick6: yarrow:/rhs/brick1/metadata Brick7: zod:/rhs/brick2/metadata Brick8: yarrow:/rhs/brick2/metadata Brick9: zod:/rhs/brick3/metadata Brick10: yarrow:/rhs/brick3/metadata Brick11: zod:/rhs/brick4/metadata Brick12: yarrow:/rhs/brick4/metadata Brick13: zod:/rhs/brick5/metadata Brick14: yarrow:/rhs/brick5/metadata Brick15: yarrow:/rhs/brick6/metadata Brick16: zod:/rhs/brick6/metadata Options Reconfigured: cluster.tier-mode: cache features.ctr-enabled: on performance.readdir-ahead: on Also, file promotes and demotes are happening(sanity check)only when the watermark criteria is hit. changed the mode to test and test mode works as previous default behavior Hence moving to verified
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://rhn.redhat.com/errata/RHBA-2016-0193.html