freeze thaw state and response

[crpb]

I usually do freeze, flush , edit, thaw. But thats something else :p

gromox/tools/mbop_main.cpp

Lines 195 to 214 in cf051aa

	static int freeze_main(int argc, char **argv)
	{
	if (HX_getopt6(g_options_table, argc, argv, nullptr,
	HXOPT_USAGEONERR) != HXOPT_ERR_SUCCESS || g_exit_after_optparse)
	return EXIT_PARAM;
	/*
	* db_maint_mode::hold is not offered presently, since exmdb is prone
	* to get into out-of-memory situations when used.
	*/
	enum db_maint_mode mode = g_fast ? db_maint_mode::reject :
	db_maint_mode::reject_waitforexcl;
	auto ok = exmdb_client->set_maintenance(g_storedir, static_cast<uint32_t>(mode));
	if (!ok) {
	mbop_fprintf(stderr, "%s: the operation failed\n", argv[0]);
	return EXIT_FAILURE;
	}
	return EXIT_SUCCESS;
	}
	}

gromox/tools/mbop_main.cpp

Lines 167 to 168 in cf051aa

	else if (strcmp(argv[0], "thaw") == 0)
	ok = exmdb_client->set_maintenance(g_storedir, static_cast<uint32_t>(db_maint_mode::usable));

Hey, sure i can guess that the return code zero will mean that those calls worked but maybe this -v switch could make it tell others than itself.
And might it be possible to somehow retrieve the status?

Here i want to freeze snap, thaw. I do that but does it make a difference and can i somehow check if it would have even made a difference or maybe i should have asked a little nicer to freeze so i can begin with my stuff 🙈

o/

Christopher

[crpb]

So i found an easy way to detect it

gromox-mbop -v foreach.mb.here freeze
grommunio-index -Aq

as grommunio-index has no handling at all for that other than.

[FATAL] exmdb call failed: Dispatch error
        ... exited with status 2

Maybe grommunio-index should get some kind of timeout for that or at least should only exit with a warning as "freeze" most likely is only for a short time.

[jengelh]

I have no idea what you want. Maybe stick to the classic "what I did; what I expected; what I observed" bug report pattern instead.

[deajan]

@crpb Count me in for an answer,
I have Grommunio VMs which I quiesce in order to make backups. The quiescing executes a freeze/thaw script which at the time flushes mysql tables with a readlock and unlocking them on thaw.
I'd love to add a gromox freeze/thaw process.
Btw, having a "global" freeze/thaw instead of a foreach .mb.here one would be nice too.

[crpb]

"what I did; what I expected; what I observed"

well, that this doesn't take ages to let me unload it.
sure clients are still trying to connect to the mailbox but if i freeze it the docs say it will only let the current actions finish but not let anyone else in anymore.

how long should this take until i can actually unload the mailbox so i can do an actual filesystem freeze to not even bother with wal's as i've seen that those had issues on recovery. and to exclude this possibility i wanted to be sure.
and i didn't want to create additional sqlite database backups that then would need to be merged back into the folders and excluded during rsync or whatever.

i expected it to maybe take a few seconds at most 30!? seconds but it doesn't seem like it.

gromox-mbop -u cb@clownflare.de freeze
( while true ; do sleep 5; echo try unloading; gromox-mbop -u cb@clownflare.de unload 2>&1 && break || true ; done ) |ts "%F %H:%M:%.S"

or must i reload-restart http or something else to accomplish it?!

fwiw, the thingy i use to backup my shit where i manage the zfs-based storage aswell (hence zfs-snapshot takes care of the history)
https://github.com/crpb/grommunio/blob/main/tools/grombak.xfs

[jengelh]

When a client which listening for events on a folder, that too counts as an active operation (and it's one that will not finish by design), which can currently prevent unload from completing.

gromox-mbop(8) documents that:

unload will fail to succeed if there is still a client connected to the mailbox via a notification channel.

[crpb]

And this leads to what operation? Reload one of the services as described in https://github.com/grommunio/gromox/wiki/Reloading and it will kill it? Or do i need to take the chainsaw called systemctl restart nginx gromox-http?
Haven't seen any tool to list those particular connections and "manage" them :P

EDIT: oh, and this will also mean as long as that one connection doesn't finish it could still do writes or not? so we will never see the mailbox freeze over