How My Mileage Varied September 25, 2009
Posted by Matt Cook in Domino.trackback
When reviewing Planet Lotus recently, I came across this Stephan Wissel entry re: minimizing your server downtime when upgrading Domino.
As we had several sites with server lease swaps pending, I decided to try it out and test the procedure.
Typically at the smaller sites, I will take down the current server, copy the data over into the Domino directories, update notes.ini, install anti-virus, etc. My plan was to get all of the data moved during the week to hopefully only work an hour or so on the weekend.
I appreciated not having to wait for a file copy and that Domino wasn’t down for long to do that copy.
Copying the notes.ini lines over from the old to the new server was tedious and my least favorite part of the procedure.
What ruined my weekend and the following few work days were two gotchas.
The first was that files were moved via adminp from a 6.5.5 FP3 server to an 8.5 FP1 server. There is a known issue with moving mail files via adminp and unread marks not replicating. I should have paid better attention to Kathleen McGivney’s admin4 session at IamLUG as this was documented here. I also should have had someone at the site test the mail files before we proceeded with migration/swap. I assumed all would be OK and suffered the results.
The second issue was in creating replicas of several applications with replication formulas, the formulas were lost as they were configured when servername replicates with any server. Once I brought up the migration server as the old server I had no replication formulas in these applications. I therefore replicated an extra 15 GB of data from our hub back to a remote site over a T1 connection. These applications then had to replicate back over the T1 to the network hub and back out to sales folks connected over VPN. It was the network team that brought to my attention that all of the WAN link was consumed by Domino traffic. Ouch!
I might use this procedure again if I have a site where we can’t afford the downtime for the file copy. We can work around the unread marks and admin4 by either copying the Domino data at the OS level or upgrading the existing server to a version that has the admin4 unread mark fix. Not sure how to determine where we have replication formulas applied on applications to avoid that issue.
Hi Matt,
for the replication formula: you can use DXL to extract them and compare. This way you can verify them. And yes: some tooling to compare two NSF on 2 servers (documents, acl, replication formula) seems like a good idea to be added to a toolbox.