Deploying a large charm times out with 'openstack' provider
Bug #1110607 reported by
Sidnei da Silva
This bug report is a duplicate of:
Bug #1156649: pyJu fails to deploy/upgrade charm with 32MB payload.
Edit
Remove
This bug affects 1 person
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
pyjuju |
New
|
Undecided
|
Unassigned |
Bug Description
This may or may not be environment specific.
Deploying a charm that contains a large tarball inside (35Mb) fails with either ConnectionLost or ConnectionDone after about 30s. Seems like the server is cutting the connection short after a timeout. Smaller charms do not have an issue, and using the 'openstack_s3' provider works too.
To post a comment you must log in.
<hazmat> sounds like that's basically the server calling it kaput on the connection
<hazmat> we can do a retry.. but its not going to change the result
<hazmat> there isn't really any abstraction around the put file on our end.. read file (into mem) and PUT object with contents
<hazmat> the real question is whether the key exists after the server closes the connection
<sidnei> into mem? zomg :)
<sidnei> good question
<hazmat> sadly both txaws and our twisted openstack client suck like that
<sidnei> we have some code for streaming uploads in u1, it's a bit messy but maybe i could take a look at that
<hazmat> well i'm looking at that as well, but its not the issue
<hazmat> ie we're sending a single PUT with contents..
<hazmat> whether its streaming to the put request or not.. isn't really the question
<hazmat> how the server is behaving with the PUT request is
<sidnei> sure
<hazmat> it would be good to check the contents of the key
<hazmat> it might be a local swift config, or it might be the middleware cut off the request, but the backend processes it successfully
<sidnei> whats the best way to do that?
<hazmat> python-swiftclient
to be continued... recording for posterity