iwmigrate mystery on TS 7.4.1.1 RHEL 6.5 64bit

Binettoid

Hi all,

I'm trying to migrate the latest edition only from one store to another.

It turns out this CLT is a trickier beast than most:

I've mounted the old store on /iw-store-transfer/default as /iwmnt/old

I got absolutely nowhere for a while, using:

iwmigrate -x -x -x -s -m /iwmnt/old -o /iw-store/default -b /default/main

I took this to mean that I was passing these parameters:

-x -x -x = triple verbosity (according to the doco, which, as we'll see shortly...you know where I'm headed....;-)

-s = latest edition plus changes (thanks ghoti, tho this now causes an error saying you must use -d or -e)

-m = mount point to find the source content to migrate (now causes an error saying -m is deprecated)

-o = destination store (mustn't be a mount, since passing no -m means there can only be one store mounted)

-b = the source branch I wish to migrate

So, the errors are inline there and the command changed to this before I gave up

iwmigrate -x -x -x -s -o /iw-store/default -b /iwmnt/old/main -d -s -w

This command ran and said "Error locating edition /iwmnt/old/main/EDITION/-w"
So, does this mean we need to provide edition numbers now? Ghoti? ;-)
Ah well, we'll try -e instead tomorrow, or actually specify numbers, and post the results.

If anyone would be so kind as to confirm or correct the rest of what I posted here, I'd really appreciate it.

Cheers!

Find more posts tagged with

Comments

nipper

I thought you had to provide the starting and ending editions.

Adam Stoller

You have to provide at least the starting edition path / name.

If you're dealing with the "last" edition, that should be sufficient - if you're trying to take only one edition from the "middle", you would want to specify both '-s' and '-e', but I think that's a very rare situation.

Binettoid

Thanks Nipper and Ghoti for your responses on this.

Old hands like you guys may still find your way with this tool and there are some use-cases that seem simpler and should be easy enough. In our case, despite the tool reporting success, it copied e.g. "main" inside the unmounted destination iw-store/default. Needless to say, this didn't work once the store was activated.

It seems they have changed the tool so that it's easier to use for implementations with only one store license. Does that make sense? However, in my opinion this CLT is different enough from its documentation, which still lists the -m flag as a required parameter, that it might as well be completely broken.

We raised a support ticket to try to get help for the correct usage of this under our circumstances and will post any light shed on this by support. Due to time constraints and the anticipated response-time of support, we'll be migrating our entire store and using a scripted iwrmed operation to remove all preceding editions of each branch.

I'm guessing (well, it isn't documented ;-) that we'll need to keep INITIAL as well as the latest editions. Any advice here is much appreciated.

Adam Stoller

Binettoid wrote:
[...]. Due to time constraints and the anticipated response-time of support, we'll be migrating our entire store and using a scripted iwrmed operation to remove all preceding editions of each branch.
I'm guessing (well, it isn't documented ;-) that we'll need to keep INITIAL as well as the latest editions. Any advice here is much appreciated.

I strongly advise against this strategy - iwrmed will cause LOTS of load on your server for an extended period of time (hours, days, weeks, or even months!)

If you still decide you have to do it this way - delete one edition on one branch at a time and run iwstat -c -i sleep_interval to monitor the process of deleting the edition, so that you don't start deleting the next edition until the previous one has been deleted (in this case 'next' and 'previous' refer to individual editions, not whether one edition precedes or follows another). As you can guess - this will take a long time to complete - but you will keep the load on your server at a reasonable level throughout the process. Again, I'd strongly advise not trying to delete more than one edition at a time; it may seem to be the logical thing to do (especially overnight), but you are more likely to run into problems where the thread processes are running way longer than you expected and the load on the server starts impacting users.

nipper

Echoing Ghoti's comments. I have seen rmed take several weeks,

Binettoid

Hey guys thanks for the tips.

Luckily we are in a pre-production phase, so there are no users accessing the environment yet and anyway we try things on a VM mirror of the working environment first, of course. As a result, and before you warned against this, we were able to test removing some editions in a few different ways:

Removing them manually via the UI - first a couple, then 100 at a time.

Later, scripted removal of 500 at a time.

Also, removal of an entire branch with 3,500 editions via the UI.

Subesquently, we ran iwfsshrink.

There were indeed significant spikes of CPU and throughput activity for some time but these usually receded after 10 to 15minutes. Our real server has significantly more CPU, memory and faster raid SSD drives, etc. so we'll be using the same techniques on the real environment.

N.B. it may be worth considering that our TeamSite usage pattern is, in my experience, not very 'heavy', and editions typically contain ~200mb of static files with little change between editions. Also there is no datadeploy activity and this amy all help to make these operations less impactful. It would be interesting to see how this goes on more 'advanced' setups using extensive DD, etc.