Re: question regarding 5.0.8 -> 5.0.11 upgrade

[Keith Fong <keith@sfu.ca>]

We use Zimbra's HTTP/IMAP/POP Proxy on two servers, and use our F5 to load balance/failover between the two.  This allows us to publish one URL for our 4 mailboxes (we are currently on 5.0.11).

One suggestion I have is to remove the tcp timeouts from the proxy pools on the load balancer, and allow nginx to manage it.  We had some unexpected behaviour with certain IMAP clients when the timeouts on the load balancer were shorter than the ones specified in nginx.


Keith


--
Keith Fong
Systems Consultant
Institutional, Collaborative, and Academic Technologies (ICAT)
IT Services
Simon Fraser University

----- Original Message -----
From: "Tom Golson" <tgolson@tamu.edu>
To: "zimbra-hied-admins" <zimbra-hied-admins@sfu.ca>
Sent: Tuesday, March 24, 2009 2:21:20 PM GMT -08:00 US/Canada Pacific
Subject: Re: question regarding 5.0.8 -> 5.0.11 upgrade

Somewhat off-topic, but do you mind me asking how your mail service is
presented to your customers?  We haven't enabled any of the proxying
that's generally available, now.  We essentially round-robin the service
behind a pair of F5's and then rely on the redirect behavior of Zimbra
if you log into the wrong mailstore.

We were looking at the 5.0.15 update but this bug
http://bugzilla.zimbra.com/show_bug.cgi?id=35893 made us a little
nervous.  Have you looked at this?

--
Tom Golson
Chief Systems Engineer
Open Systems Group
Computing & Information Services
Texas A&M University

Xueshan Feng wrote:
> ----- "Michael A. Jastremski" <jastrems@isc.upenn.edu> wrote:
> 
> | Good afternoon,
> | 
> | We are planning an upgrade of our multi-server environment in the next
> | month or so.
> | I'm writing today to ask a few questions ..
> | 
> | - Have you upgraded to 5.0.11 from a prior 5.0.X release?  If so, was
> | this a multi-server environment?  
> 
> Yes in anmulti-server environment, with 3 ldap servers, 8 production mailbox servers, 2 spares.
> 
> | - How long did the process take?  How much time was initially
> | allocated? 
> 
> 
> Allocate 2 hours, usually takes 1.5 hours.
> 
> | - Any new problems with 5.0.11 ?
> 
> Why not go to 5.0.15?
> 
> We upgrade in test environment first ( 2 ldap servers, 2 mailbox servers). No problem found for the upgrade process.
> 5.0.15 production system upgrade is scheduled this coming weekend from 5.0.12 to 5.0.15.
> 
> 
> | - How many users/gigabytes were involved in your upgrade?  
> 
> Average 5000 users per server, 1.5 TB used space in /opt/zimbra/store.
> 
> 
> | - Anything you'd do differently a second time around?
> 
> We've done it 10 times over (starting from 5.0.3).  Most upgrade and post-upgrade are managed by configuration management ( things that would not be kept by upgrade, such as
> skin, java policy if you use kerberos) and  scripts  (disable/enabling monitoring,  put up outage page on webmail, post-upgrade tests etc).  
> 
> Xueshan
> 
> | 
> | Thank you very much in advance.
> | 
> | 
> | _Michael.
> | 
> | 
> | --
> | Michael Jastremski
> | Systems Programmer
> | ISC Networking
> | University of Pennsylvania