REDIS HA

23 views
Skip to first unread message

Eric Fetzer

unread,
Aug 25, 2021, 9:33:44 AMAug 25
to Redis DB
Hi all!  So I'm planning my architecture of REDIS HA.  The primary goal is to be able to do patches on servers one at a time without taking an outage.  Here's my plan:

RHEL 7.9 Server 1:
Primary REDIS-Server
1 Sentinel

RHEL 7.9 Server 2:
Replica REDIS-Server
1 Sentinel

RHEL 7.9 Server 3:
1 Sentinel

I plan to run the REDIS Servers and Sentinel instances as SystemD services.  I'll have them auto start on boot starting the Server first and then having the Sentinel dependent on the Server.  Obviously the Sentinel server will have no dependency.  For patching, we would take down one server at a time, patch, and bring it up.  Does anyone see issues with my architecture?

Thanks,
Eric

Reynold Yick

unread,
Sep 1, 2021, 6:49:41 AMSep 1
to Redis DB
Hi Eric,

I am using similar architecture with latest version 6.2.5, unfortunately I hadn't worked on patching at this moment.

1. a F5 load balancer

2. Redis Primary Node: RHEL 7.9 + redis + redis-sentinel, configured with
2.1 "replica-priority 10" to work as primary
2.2 "sentinel down-after-milliseconds mymaster 2000"
2.3 "sentinel failover-timeout mymaster 9000"

3. Redis Secondary Node 1: RHEL 7.9 + redis + redis-sentinel, configured with
3.1 "replica-priority 100" to work as standby
3.2 "sentinel down-after-milliseconds mymaster 2000"
3.3 "sentinel failover-timeout mymaster 9000"

4. Redis Secondary Node 2: RHEL 7.9 + redis + redis-sentinel, configured with
4.1 "replica-priority 0" to work on replication only
4.2 "sentinel down-after-milliseconds mymaster 2000"
4.3 "sentinel failover-timeout mymaster 9000"

Please be noted that the above architecture is still encountering 5 to 10 seconds outage due to switch over master role to node 2. This period consumed with load balancer detection and redis-sentinel (role) switch over. If you wish to have full HA architecture, I believe you should consider to implement with redis cluster. Thank you.

Eric Fetzer

unread,
Sep 1, 2021, 11:29:17 AMSep 1
to redi...@googlegroups.com
Thanks Reynold, 10 second outages are acceptable for what we're storing in REDIS.  Your architecture is what I hope to one day move to with the third full node.  Right now I'm just borrowing from another server in the domain to come up with the third sentinel to reach quorum.

--
You received this message because you are subscribed to a topic in the Google Groups "Redis DB" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/redis-db/SdWCoAB72cU/unsubscribe.
To unsubscribe from this group and all its topics, send an email to redis-db+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/redis-db/b44c44f8-cebd-4e6d-8248-cca5bc34d52bn%40googlegroups.com.
Reply all
Reply to author
Forward
0 new messages