Useful advice for OOM during smart server command?

[Martin (gzlist)]

What's most practical suggestion to give to a user who hits an out of
memory situation, particularly when the problem occurs on a remote
server? I'm looking to add some documentation[1] that's more helpful
than generic "upgrade bzr" or "buy more ram" responses.

It's obviously a condition users run into in practice, and not all of
them of the Don't Do That Then "trying to commit my 4GB raw video"
kind of thing. Though unlike other projects[2], bzr doesn't really
have a good story for those people.

A number of the bug reports seem to be during auto-repacking[3][4],
such as a repo with a 68MB file on a vm with 256MB memory allocated[5]
during push.

Possible options I've seen suggested in the past:

* Rewrite history to omit the giant file
* Retry the operation with nosmart+
* Run `bzr pack nosmart+<url>`

The first is probably the best solution in many circumstances, but
isn't trivial and I'm not sure what existing documentation we have for
the process.

Martin


[1]: "Proposal to merge branch"
https://code.launchpad.net/~gz/bzr/smarter_smart_error_reporting_full_722416/+merge/51969
[2]: "Large file management with git-annex"
http://lwn.net/Articles/419241/
[3]: "Out of memory error in _ensure_content on auto repacking"
https://bugs.launchpad.net/bzr/+bug/602614
[4]: "AssertionError in
bzrlib._groupcompress_pyx.DeltaIndex._populate_first_index"
https://bugs.launchpad.net/bzr/+bug/633336
[5]: "out of memory in
bzrlib._groupcompress_pyx.DeltaIndex._populate_first_index"
https://bugs.launchpad.net/bzr/+bug/721897