Opened at 2008-02-14T01:04:57Z
Closed at 2009-02-25T03:47:39Z
#315 closed defect (fixed)
tahoe restart barfs on non-node directories
Reported by: | zandr | Owned by: | warner |
---|---|---|---|
Priority: | minor | Milestone: | 1.4.1 |
Component: | code-nodeadmin | Version: | 0.8.0 |
Keywords: | Cc: | ||
Launchpad Bug: |
Description
Allmydata.com has a convention of having all operational nodes in a directory called ~/prodnet. This makes restarting all nodes easy, doing something like "for i in *; do tahoe restart $i; done" (or even tahoe restart -m *).
Unfortunately, there are two things that belong in ~/prodnet that aren't actually nodes. (stats-gatherer and log-gatherer) These cause problems because while tahoe start and tahoe stop will ignore them, if tahoe restart encounters an error during the stop, it won't do the start.
tahoe restart should ignore "not a node directory" errors, at least until the stats and log gatherers can be wrapped up in something that can be controlled by tahoe start.
Change History (9)
comment:1 follow-up: ↓ 3 Changed at 2008-02-14T03:22:18Z by zooko
comment:2 Changed at 2008-02-14T03:22:45Z by zooko
- Owner set to zandr
comment:3 in reply to: ↑ 1 Changed at 2008-03-05T00:41:47Z by zandr
- Owner changed from zandr to zooko
- Version changed from 0.7.0 to 0.8.0
Replying to zooko:
? This will go ahead and restart all the things in prodnet that are nodes, right? And the presence of non-nodes won't prevent the restart of the nodes?
That will work, but tahoe restart -m ~/prodnet/* is tripped up by the presence of non-nodes. It will stop everything and start nothing if it finds a non-node.
The real fix for this is to get the stats-gatherer and flogtool running in containers that can be controlled by tahoe [re]start. I thought a ticket existed on that, but I couldn't locate it so I created a new one, #330 , which I pointed at warner.
comment:4 Changed at 2008-03-25T18:40:09Z by zooko
- Milestone changed from 1.0.0 to 1.0.1
comment:5 Changed at 2008-03-25T18:52:42Z by warner
FYI, the new foolscap-0.2.5 (released last night) provides a 'flogtool create-gatherer' command, which creates a working directory with a .tac file, just like 'tahoe create-client'. This working directory can be launched with 'tahoe start' and shut down with 'tahoe stop'.
I plan to make the tahoe stats-gatherer behave the same way. I also intend to fix tahoe restart to tolerate shutdown failures.
comment:6 Changed at 2008-05-05T21:08:36Z by zooko
- Milestone changed from 1.0.1 to 1.1.0
Milestone 1.0.1 deleted
comment:7 Changed at 2008-05-29T22:26:48Z by warner
- Milestone changed from 1.1.0 to 1.2.0
comment:8 Changed at 2008-05-29T22:27:07Z by warner
- Owner changed from zooko to warner
comment:9 Changed at 2009-02-25T03:47:39Z by warner
- Milestone changed from 1.4.0 to 1.3.1
- Resolution set to fixed
- Status changed from new to closed
fdb0200ecad291e2 changes 'tahoe restart' to use --force by default (#165), which will be in 1.3.1 . The stats-gatherer was turned into a regular 'tahoe start'-able node in 1.3.0, and the foolscap log-gatherer was converted in foolscap-0.3.0 . So I think we're done with this ticket.
I don't understand the problem -- you mean
? This will go ahead and restart all the things in prodnet that are nodes, right? And the presence of non-nodes won't prevent the restart of the nodes?
Also, as per #165 -- "restart should start if the node isn't running" -- you can add -f or --force to cause restart to start the node even if it can't stop the node first.