questions about api/stats/query results

23 views
Skip to first unread message

Huafeng Lyu

unread,
Mar 21, 2017, 7:25:49 AM3/21/17
to OpenTSDB
Hi, I have some questions about the results of the api/stats/query api. I am using opentsdb 2.3.


something strange:
(1) some 'max' values are smaller than their 'avg' counterparts: aggregationTime, queryScanTime, serializationTime. Why?
(2) maxScannerTime is not in the result. Why?
(3) avgUidToStringTime is not in the result. Why?
(4) The found the following explanation about scannerTime and queryScanTime from opentsdb documentation:
scannerTime: The total time from initialization of the scanner to when the scanner completed and closed.
queryScanTime: Total time spent waiting on the scanners to return results. This includes the groupByTime.
So it looks like scannerTime should be greater than queryScanTime ( I guess: scannerTime = scanner_init_time + queryScanTime + some_other_time).
But the result shows scannerTime < queryScanTime a lot.

Can anyone help me understand these? Thanks.
--huafeng

ManOLamancha

unread,
Apr 25, 2017, 4:38:45 PM4/25/17
to OpenTSDB
On Tuesday, March 21, 2017 at 4:25:49 AM UTC-7, Huafeng Lyu wrote:
Hi, I have some questions about the results of the api/stats/query api. I am using opentsdb 2.3.


something strange:
(1) some 'max' values are smaller than their 'avg' counterparts: aggregationTime, queryScanTime, serializationTime. Why?

Bugs. I have to re-work the tracing code.
 
(2) maxScannerTime is not in the result. Why?

Another bug, must have missed it :( 

(3) avgUidToStringTime is not in the result. Why?

Another bug :(
 
(4) The found the following explanation about scannerTime and queryScanTime from opentsdb documentation:
scannerTime: The total time from initialization of the scanner to when the scanner completed and closed.
queryScanTime: Total time spent waiting on the scanners to return results. This includes the groupByTime.
So it looks like scannerTime should be greater than queryScanTime ( I guess: scannerTime = scanner_init_time + queryScanTime + some_other_time).
But the result shows scannerTime < queryScanTime a lot.

It's worded poorly but the queryScanTime is the total time spent waiting for all of the scanners to complete. It's the "scanner init + (max(salt scanner time))" so query scan time is always greater. 
Reply all
Reply to author
Forward
0 new messages