Nginx Bad Gateway occurs when one mailbox server is unavailable

[Jesster]

Hey All,

We have a 5.0.13 Multi Server Install:

2 Proxies
2 Mailbox Servers
2 LDAP Servers
2 MTA Relay Servers
12 MTA MX & AntiVirus/AntiSpam Servers


The two proxies handle all reverse-proxied pop,imap, http/https traffic for Zimbra to our users.
When we take one of the mailbox servers down, the Proxies immediately return 502 Bad Gateway for http/https users.

What's the best way to take a mailbox server down without nginx disrupting service?

$ zmprov garpu
zmbx01.example.org:7072/service/extension/nginx-lookup zmbx02.example.org:7072/service/extension/nginx-lookup

$ zmprov gamcs
zproxy01.example.org:11211 zproxy02.example.org:11211

Thanks!

-Jesster

[Klug]

When we take one of the mailbox servers down, the Proxies immediately return 503 Bad Gateway for http/https users.

You're getting this whatever mailbox server you put down ?

Or you're getting this only when a user on the server you put down wants to log in ?

[Jesster]

You're getting this whatever mailbox server you put down ?

Or you're getting this only when a user on the server you put down wants to log in ?

Klug,
I had to test this to be sure. I logged myself out of Webmail, waited about 2 minutes, then shut zimbra down on zmbx02, sure enough, I got "502 Bad Gateway".

[Jesster]

It seems our users experience this even when both mailbox servers are online. We cannot identify patterns by the impaired users. The problem is intermittent .

I've searched the forums, and found posts about dns, host files, and zimbraMailReferMode on the mailbox servers.

Perhaps it's time to open a Support Case..

[mmorse]

-Might set zmprov mcf zimbraReverseProxyLogLevel debug for a bit.
-nginx logs (when it happens / if you get a chance to trigger it)
-nginx conf files
-zmprov gs `zmhostname` | grep Proxy (on proxies & stores, or better yet without the grep if your sending a file to support - label which is which what your passing `zmhostname` instead of ztype0#)