Closed Bug 518271 Opened 15 years ago Closed 15 years ago

make node assignment smarter and support delayed assignment to a node

Categories

(Cloud Services Graveyard :: Server: Sync, defect, P1)

defect

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: mconnor, Assigned: telliott)

References

Details

(Whiteboard: [needs production push])

Instead of sticking everyone onto whatever the currently-specified node, we should distribute evenly up to a per-node cap.  We should also cap new node assignments to better avoid spikes from new user setups.  I'll leave it to Toby/zandr to figure out what the user/signup caps should be, but if we can't assign a node to a user, we should just return an empty string and the client will deal.

Ideally, chknode should attempt to find the user a home on subsequent queries, since that'll give us complete control over new user spikes...

I'll file a bug on the client side of this (and implement it, since it doesn't depend on the server side at all).
Flags: blocking-weave1.0+
Depends on: 518273
This is fixed on weave-dev, will push to production in Tuesday's push.
Status: NEW → ASSIGNED
Whiteboard: [needs production push]
Target Milestone: 0.7 → 0.8
Toby, what's left here?
The stuff I emailed about last week re the dashboard. Since nobody had any objections to the plan, I'll file the bugs.
Depends on: 524584
No longer depends on: 518273
Depends on: 524586
Depends on: 524585
Once the blockers are done, this is enabled by setting WEAVE_REGISTRATION_THROTTLE_DB to any value, and defining WEAVE_REGISTRATION_THROTTLE_HOST, WEAVE_REGISTRATION_THROTTLE_DB, WEAVE_REGISTRATION_THROTTLE_USER and WEAVE_REGISTRATION_THROTTLE_PASS

Let's not do that while we're not all around to test and watch for problems...
Whoops, ignore the dupe constant above
Status: ASSIGNED → RESOLVED
Closed: 15 years ago
Resolution: --- → FIXED
Product: Cloud Services → Cloud Services Graveyard
You need to log in before you can comment on or make changes to this bug.