Gerrit HA Design with HA Proxy server

Bharat Darshi

unread,

Jul 11, 2024, 1:26:40 AM (11 days ago) Jul 11

to Repo and Gerrit Discussion

Hi All,

We have designed a HA Gerrit setup of 2 Gerrit servers with a HA Proxy server that is routing traffic in a round robin fashion. This has been done in AWS. Please let us know your views if this architecture is good for handling high traffic load of around 100-200 users who would use Gerrit heavily.

Our setup will have provision for autoscaling and a replica HA server will also be added in the future.

Thanks,

Bharat Darshi

Bharat Rajasekhar

unread,

Jul 11, 2024, 1:40:30 AM (11 days ago) Jul 11

to Repo and Gerrit Discussion

--
--
To unsubscribe, email repo-discuss...@googlegroups.com
More info at http://groups.google.com/group/repo-discuss?hl=en

---
You received this message because you are subscribed to the Google Groups "Repo and Gerrit Discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to repo-discuss...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/repo-discuss/44794c7e-2e40-44cf-8275-3e5987e5f486n%40googlegroups.com.

IMG_20240711_110915657.jpg

Matthias Sohn

unread,

Jul 11, 2024, 3:01:17 AM (11 days ago) Jul 11

to Bharat Darshi, Repo and Gerrit Discussion

On Thu, Jul 11, 2024 at 7:26 AM Bharat Darshi <bharatra...@gmail.com> wrote:

Hi All,

We have designed a HA Gerrit setup of 2 Gerrit servers with a HA Proxy server that is routing traffic in a round robin fashion. This has been done in AWS. Please let us know your views if this architecture is good for handling high traffic load of around 100-200 users who would use Gerrit heavily.

I think you should consider adding the high-availability plugin [1] to ensure synchronisation between the servers.

The healthcheck plugin [2] can be used to let haproxy react on the health status of the 2 servers.

This should be good for an active/passive setup.

To support active/active you can add the global-refdb module [3] and its AWS implementation aws-dynamodb-refdb [4]

to prevent split brain issues.

Another option is k8s-gerrit [5] which provides an HA setup on kubernetes supporting multiple Gerrit primary and replica pods.

[1] https://gerrit.googlesource.com/plugins/high-availability/+/refs/heads/master

[2] https://gerrit.googlesource.com/plugins/healthcheck/+/refs/heads/master

[3] https://gerrit.googlesource.com/modules/global-refdb/+/refs/heads/master

[4] https://gerrit.googlesource.com/plugins/aws-dynamodb-refdb/+/refs/heads/master

[5] https://gerrit.googlesource.com/k8s-gerrit/+/refs/heads/master

-Matthias

Our setup will have provision for autoscaling and a replica HA server will also be added in the future.

Thanks,
Bharat Darshi

--

Reply all

Reply to author

Forward