676296 - Production push request for server-node-assignment rpm-1.0-1

Status: VERIFIED FIXED

(Cloud Services :: Operations: Deployment Requests - DEPRECATED, task)

Description

[Toby Elliott [:telliott]]

Attachment: SQL statements for altering the nodes table.

This is the new server node assignment. Requests to user/node/weave should redirect to it (and ideally be rewritten to user/node/sync)

QA plan:

Make sure new nodes are being assigned from this correctly. Feel free to coordinate with operations to set up limited availability nodes and do requests from different users. Same users should stay persistently on a node.

Ops notes: requires SQL changes (attached)

Comment 1

[Toby Elliott [:telliott]]

Attachment: Revised SQL statements for altering the nodes table

Adds the requested migrating column

Comment 2

[Pete Fritchman [:petef]]

Build problems:
- missing python26-wsgiproxy package (dependency also missing)

Other problems:
- missing a run.py to run through gunicorn. It should load /etc/mozilla-services/nodes/production.ini.

Should I be able to fetch http://<node server>/user/1.0/<user>/node/weave?

I get a 404 when I do that (with a valid user).  I get a 503 when I hit "/1.0/<user>/node/weave", and the stack trace (which only shows up in stdout, not my /var/log/nodes.log), looks like:

@400000004e4495e6294f1c84   File "/usr/lib/python2.6/site-packages/nodes/controller.py", line 73, in assign_node
@400000004e4495e6294f283c     self.auth.get_user_info(request.user, ['syncNode'])
@400000004e4495e6294f300c TypeError: get_user_info() takes exactly 2 arguments (3 given)

Comment 3

[Toby Elliott [:telliott]]

(In reply to Pete Fritchman [:petef] from comment #2)

> Should I be able to fetch http://<node server>/user/1.0/<user>/node/weave?

No. The valid URL is /1.0/<user>/node/sync - we'll be using redirect magic to get there.

Will fix the other things now.

Comment 4

[Toby Elliott [:telliott]]

It's now rpm-1.1-1 and is in the MozSvcNodes namespace

Comment 5

[Pete Fritchman [:petef]]

not ready for deploy now; ldap authentication/api spec issues.

Comment 6

[Toby Elliott [:telliott]]

Attachment: Updated alter to reflect larger node column

Comment 7

[Toby Elliott [:telliott]]

Reopening as it's going on this week's train. Turns out that we already have node-sreg, so it's just a matter of having the VMs ready.

Comment 8

[James Bonacci [:jbonacci]]

So, does this mean that https://bugzilla.mozilla.org/show_bug.cgi?id=678579 is back in as well?

Comment 9

[Toby Elliott [:telliott]]

It's not dependent. Basically, once the nodes launch, the functionality in sreg becomes redundant, but there's no need to pull it immediately. 

If we had to roll back the nodes deployment, that would also mean having to roll back the sreg deployment, and that seems like unnecessary work at this point.

Comment 10

[Pete Fritchman [:petef]]

Deploy note:  need an updated weave_acls.conf with snodes IPs.

Comment 11

[Pete Fritchman [:petef]]

Latest rpm version in stage: rpm-1.3-1

Comment 12

[Pete Fritchman [:petef]]

shipped