[SERVER-9104] currentOp output format changed for MapReduce Created: 23/Mar/13  Updated: 10/Dec/14  Resolved: 08/May/13

Status: Closed
Project: Core Server
Component/s: Diagnostics, MapReduce
Affects Version/s: 2.4.0
Fix Version/s: None

Type: Bug Priority: Major - P3
Reporter: Anton Sviridov Assignee: Daniel Pasette (Inactive)
Resolution: Done Votes: 0
Labels: currentOp
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Operating System: Linux
Participants:

 Description   

I had an ugly script that was launching several MapReduce tasks consequently and handled the execution progress in each of them to show nice messages like "done: 75%" and ensure that no MapReduce task will be executed concurrently with another reducing performance of both.

It relied heavily on the output of currentOp(or, to be precise, on the contents on $cmd.sys.inprog collection), using "query" parameter to find command in the list of currently executing commands. However, when I updated to 2.4.0, db.currentOp shows

"query" :

{ "$msg" : "query not recording (too large)" }

,
for all my MapReduce tasks, thus breaking my observer script and laying raid to my cattle. Has there been any logic behind hiding query information in currentOp?



 Comments   
Comment by Daniel Pasette (Inactive) [ 08/May/13 ]

Hi Anton, as this format has not changed since before version 2.0, I'm closing this out. I hope using either state saved in your client app or in the db itself should suffice for controlling your job execution.
Thanks
Dan

Comment by Anton Sviridov [ 27/Mar/13 ]

Dan, I am using mongo on my ubuntu desktop, so updates are received automatically and I don't really keep track of the versions I've been using. 2.4 is the first mongo update I've received so far, I guess it was 2.2.something.

I did try parallel M/R, yes. It is unfortunately of no use in my pipeline based system, but I did ask a question on mongo blog about whether this new functionality might help with emulating multithreaded M/R on a single node. So I'm looking forward to employing this new feature.

Comment by Daniel Pasette (Inactive) [ 24/Mar/13 ]

Can you tell me what version were you upgrading from? I am fairly certain currentOp() output has truncated the query field to <= 512 bytes for many versions. On a related note, have you tried running map/reduce jobs in parallel with 2.4 to take advantage of the new v8 javascript engine?

Generated at Thu Feb 08 03:19:24 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.