Administrators • Error in url loginOp=logout during rolling upgrade

Hello

we have an environment where ldap and proxy node are already migrated to 10.1, 9.x store node are hosting the mailbox and 10.1 store nodes has been provisioned.

Before starting user migration zimbraReverseProxyUpstreamLoginServers has been configured with only the 10.1 servers.

We have a strange behaviour where users who are still using "classic" on the 9.x servers get a 502 error for the URL "?loginOp=logout" when they log out of the webmail, whereas for those migrated and using "classic" on 10.1, the logout works correctly.

Relevant logs at the proxy level:

Dec 9 10:58:18 umana-z10-proxy01 zimbra: 2024/12/09 10:58:18 [error] 54960#0: *1085567 no live upstreams while connecting to upstream, client: xx.xx.xx, server: mailstore01 request: "GET /?loginOp=logout HTTP/2.0", upstream: "https://zimbra_login_ssl/?loginOp=logout", host: "xx.xx.xx", referrer: "https://xx.xx.xx/"

In Nginx upstream zimbra_login_ssl is populated with the 10.1 servers.

This configuration block handles the logout URLs:

# if to loginOp is logout, all the requests should be routed to login server so that updated assets can be loaded
if ($arg_loginOp = "logout") {
set $login_upstream https://zimbra_login_ssl;
}

Has anyone noticed this behavior during the migration phase?

Thanks.

Statistics: Posted by dapa6 — Tue Dec 10, 2024 8:30 pm

---