a single memcache host should be used for keystone cache

Bug #1471264 reported by Boris Bobrov
8
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Mirantis OpenStack
Won't Fix
Wishlist
MOS Keystone
6.1.x
Won't Fix
High
Registry Administrators
7.0.x
Won't Fix
High
Boris Bobrov
8.0.x
Won't Fix
Wishlist
MOS Keystone

Bug Description

Now fuel sets option [cache]memcache_servers in keystone.conf to a list of memcache hosts, each host being a controller. For example, if controllers have ips 192.168.0.6, 192.168.0.7 and 192.168.0.8, then [cache]memcache_servers=192.168.0.6,192.168.0.7,192.168.0.8 on all controllers. This creates an unnecessary point of care about HA. If a node goes down, we need to wait at least socket_timeout seconds for each cache check.

[cache]memcache_servers should be set to a single local memcache instance: [cache]memcache_servers=192.168.0.6 for controller with ip 192.168.0.6 etc.

Pros:
 - faster cache -- no need to connect somewhere via network
 - no need to care about HA in cache

Cons:
 - cache is no longer shared

Revision history for this message
Boris Bobrov (bbobrov) wrote :

I would also like to have this backported to 6.1

Changed in mos:
importance: Undecided → Medium
status: New → Confirmed
Boris Bobrov (bbobrov)
tags: added: ha
Revision history for this message
Oleksiy Molchanov (omolchanov) wrote :

Need to have scale test results in order to compare and make a decision.

Changed in mos:
status: Confirmed → Won't Fix
status: Won't Fix → Incomplete
Revision history for this message
Bogdan Dobrelya (bogdando) wrote :

The 7.0 will use fernet tokens from the Kilo. Hence, invalid for the 7.0

Changed in mos:
status: Incomplete → Invalid
Revision history for this message
Boris Bobrov (bbobrov) wrote :

No, it is valid for 7.0. This is about cache, not about tokens.

Changed in mos:
status: Invalid → Incomplete
Revision history for this message
Boris Bobrov (bbobrov) wrote :

for some reason scale tests fail with message that user X has no access to tenant Y. I am investigating it now.

Changed in mos:
assignee: Fuel Library Team (fuel-library) → MOS Puppet Team (mos-puppet)
importance: Medium → High
status: Incomplete → Confirmed
Changed in mos:
assignee: MOS Puppet Team (mos-puppet) → Denis Egorenko (degorenko)
tags: added: fuel-to-mos
Revision history for this message
Denis Egorenko (degorenko) wrote :
Changed in mos:
status: Confirmed → In Progress
Revision history for this message
Bogdan Dobrelya (bogdando) wrote :

As Sergii Golovatiuk have noticed, this is not a bug, this a change request to A-A keystone architecture.

Changed in mos:
importance: High → Wishlist
Revision history for this message
Vitaly Sedelnik (vsedelnik) wrote :

Won't Fix for 6.1-updates because of Wishlist importance

Revision history for this message
Bogdan Dobrelya (bogdando) wrote :

According to review comments:
> this change only affects keystone local data cached?
yep
> Or does it affect the cached auth tokens as well
we don't cache tokens. But tokens issued and stored are not affected

this change doesn't affect HA UX and may be accepted for the 7.0 as an ordinary bugfix. Hence, returned priority back from the wishlist

Changed in mos:
importance: Wishlist → High
no longer affects: mos/8.0
Changed in mos:
status: In Progress → Fix Committed
Changed in mos:
status: Fix Committed → Fix Released
Revision history for this message
Boris Bobrov (bbobrov) wrote :

I'm going to mark this bug as "won't fix" for 7.0 due to the issues we faced (bug 1494435)

Revision history for this message
Boris Bobrov (bbobrov) wrote :

Marking as won't fix for 8.0 too for the same reason.

Boris Bobrov (bbobrov)
Changed in mos:
status: Confirmed → Won't Fix
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.