[DOCS-6721] Recommend sh.startBalancer() over sh.setBalancerState(true) for re-enabling balancing Created: 04/Dec/15  Updated: 11/Jan/17  Resolved: 03/Feb/16

Status: Closed
Project: Documentation
Component/s: manual
Affects Version/s: None
Fix Version/s: 01112017-cleanup

Type: Task Priority: Major - P3
Reporter: Spencer Brody (Inactive) Assignee: Ravind Kumar (Inactive)
Resolution: Done Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
Related
related to DOCS-1555 sh.startBalancer() might be the wrong... Closed
related to SERVER-21766 Remove waiting for balancer lock beha... Closed
is related to SERVER-22463 sh.startBalancer should be smarter ab... Closed
Participants:
Days since reply: 8 years, 2 weeks ago

 Description   

In DOCS-1555 we changed our recommendation for enabling balancing from sh.startBalancer() to sh.setBalancerState(true). In SERVER-21766, however, kevin.pulo says that support engineers still recommend sh.startBalancer(). We should have a consistent story here. Either we like the behavior sh.startBalancer() has of waiting for a balancer round to commence, in which case we should update the docs to recommend its use one again, or we don't like that behavior, in which case we should change it and still update the docs to recommend it's use. Either way we should be moving towards recommending sh.startBalancer()



 Comments   
Comment by Spencer Brody (Inactive) [ 03/Feb/16 ]

Okay, 10 points to Kay for solving the mystery! The reason sh.startBalancer() is broken is because of the balancer window. If you have a window set and you're not within the active window when you call call sh.startBalancer() or sh.stopBalancer(), those methods will block forever (well, until the window becomes active), as the balancer will not be taking the distributed lock when the window isn't active.

This is likely to all change in 3.4 with the changes we're making to how the balancer works, but for the meantime we probably shouldn't do this ticket after all and should keep the recommendations as-is for now, and re-evaluate after 3.4.

Thanks everyone who worked on this, and sorry for wasting time.

Comment by Spencer Brody (Inactive) [ 01/Feb/16 ]

ravind.kumar, I never got a response back from Richard on my last email, but it sounds like his complaints were based on either outdated or incorrect knowledge of the behavior of sh.stopBalancer(). Based on Kevin's and Andy's comments on SERVER-21766, I think we should move forward with this ticket.

Comment by Ravind Kumar (Inactive) [ 01/Feb/16 ]

spencer richard.kreuter Was there any consensus on this, since the last set of comments?

Based on reading SERVER-21766, it seems that the preference is sh.startBalancer(), but were waiting on Richard's comments.

Generated at Thu Feb 08 07:52:51 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.