Well, that'll teach me. I've been testing and testing all week. Configure new system. Do test migration, test. Do a bit more configuration. Do test migration, test. Over and over.

Got to the point today where I'm confident we should do it for real.

Did the migration work? Nope! Migration manager got stuck in a loop. Nothing useful in any log. I'm investigating what's blocking it. Might even just be a firewall issue, even though all the data got sucked up OK -- it was just failing to deploy. Luckily all these systems are virtualised to restoring to a previous point is easy. But what a horrible waste of time

strace tell you anything?

on what, though? There's all sorts of little scripts running around like mad to get migrations to do whatever it is they do It isn't as though it fails, either. It just gets stuck in some sort of loop.

I'm not sure, but it looks as though some session files it expects to find are missing.

You may know this already, but you can attack to a running process, just use this:

strace -fF -p <pid>

Poking around a little you might find the one thats "stuck" and see what its looping on.

Didn't know that, no. But there are too many threads, some python/perl and it was all very confusing.

But in the end I found it was simplest to go back three steps (images) and start from there. All went well. I still don't know what I did wrong the first time, as I reproduced my steps from my notes - maybe not in the exactl order though, so it is possible I've blocked myself from being able to migrate any more accounts to this system It could also be a munged file as as I was working from a clone of a clone of a clone.

What I'm liking is being able to run all the mod_sec rules (at least those enabled by default at any rate). But there's not been any load on the server so we'll see if we end up back with the dreaded segfaults/bus errors that I was getting in EL4 as a result of that PCRE or whatever it was bug.

From what are you trying to migrate to ? Maybe I can help. I had some pretty weird issues myself when I migrated from plesk 9 to 10.

I'm migrating grom 8.6 EL4 to 10.4.4 EL6. No problems at all. All smooth as silk thanks to extensive testing and dry-run migrations.

The only oddity was the experience I reported here, but you know what? I'm starting to doubt there was a problem at all - I think it was a refresh problem on the browser, or if it wasn't that it was a firewall issue. I don't think it was plesk's fault, at any rate.

I'm currently writing up the whole story...it will make quite a tale.

But please can you post your wierd issues? It would be helpful to know what can happen, and what to do if it does. That's a big part of my write up.