Remove duplicate metric_rollups not dealing with active relation #16166

jameswnl · 2017-10-10T19:56:31Z

https://bugzilla.redhat.com/show_bug.cgi?id=1463165

Duplicated rollup records are still there in reports.

jameswnl · 2017-10-10T19:56:45Z

@miq-bot add_label wip, bug

jameswnl · 2017-10-10T19:59:29Z

This change would fix the issue. it’s a wip coz not sure why we had that check in first place, i guess it’s for performance consideration, wanna to hold back hitting db til last.

jameswnl · 2017-10-10T19:59:52Z

@lpichler @Ladas

lpichler · 2017-10-11T08:02:59Z

app/models/metric/helper.rb

-    if recs.respond_to?(:klass) # active record relation
-      return recs unless recs.klass.kind_of?(Metric) || recs.klass.kind_of?(MetricRollup)
-    elsif recs.empty? || !recs.all? { |r| r.kind_of?(Metric) || r.kind_of?(MetricRollup) }
+    if recs.empty? || !recs.all? { |r| r.kind_of?(Metric) || r.kind_of?(MetricRollup) }


I think we can still use previous code and only change
this condition:
https://github.com/ManageIQ/manageiq/blob/master/app/models/metric/helper.rb#L124
to

return recs unless recs.klass <= Metric || recs.klass <=MetricRollup

because kind_of? is for instances and not for classes.

I think that with this change we can fix the bug and keep @kbrock's intention in #14224.

it should work also with VmPerformance because it is inherited from MetricRollup.

It would be also nice to add specs when the method has AR scope as input (MetricRollup.where(..))

thanks!

miq-bot · 2017-10-11T15:08:01Z

Checked commit jameswnl@d338c6f with ruby 2.3.3, rubocop 0.47.1, and haml-lint 0.20.0
2 files checked, 0 offenses detected
Everything looks fine. ⭐

jameswnl · 2017-10-11T16:01:53Z

@miq-bot remove_label wip

lpichler · 2017-10-11T16:36:32Z

cc @gtanzillo - just letting you know - it could affect chargeback (not sure how much when we are using max or avg)

lpichler

Looks good 👍 thanks!

Ladas

Awesome

blomquisg · 2017-10-11T20:11:14Z

it could affect chargeback (not sure how much when we are using max or avg)

@lpichler can you elaborate on your comment?

What's the overlap of max/avg and this dedup code? Are you expecting the max/avg results to change, but now be correct?

jameswnl · 2017-10-12T15:27:17Z

@miq-bot add_labels fine/yes, euwe/yes
Both fine/euwe are affected

agrare · 2017-10-13T11:45:02Z

app/models/metric/helper.rb

@@ -121,7 +121,7 @@ def self.sanitize_start_end_time(interval, interval_name, start_time, end_time)

  def self.remove_duplicate_timestamps(recs)
    if recs.respond_to?(:klass) # active record relation
-      return recs unless recs.klass.kind_of?(Metric) || recs.klass.kind_of?(MetricRollup)
+      return recs unless recs.klass <= Metric || recs.klass <= MetricRollup
    elsif recs.empty? || !recs.all? { |r| r.kind_of?(Metric) || r.kind_of?(MetricRollup) }
      return recs 


@jameswnl I just noticed this looking at the diff in git, can you remove the tailing whitespace here? It was added in 9dab1c4

agrare · 2017-10-13T11:48:07Z

@lpichler @gtanzillo it looks to me that this would make chargeback "more correct" but I understand any change in chargeback isn't good. Is this something that we should do more planning on or do you think this is okay?

lpichler · 2017-10-13T12:49:34Z

it could affect chargeback (not sure how much when we are using max or avg)
@lpichler can you elaborate on your comment?
"more correct"

yes we are now more correct but how much it is affected - it is based on how much duplicated records have been in the system for specific date range ( range the related to report whether it is monthly, weekly, daily )

Chargeback rating calculations are not related to counts of metric rollups.
Metric value (of any metric, which is used for chargeback cost calculation) is taken for the special date range and there is applied MAX or AVG.

so it can affect AVG then MAX - because for AVG we are summing up metric values and we are diving it by hours in the date range.

But previously(year before) when I was facing this issue(#11017) duplicated records mean that (duplicate records are records with same resource and timestamp)
I have one metric rollup record with any values and lot of them with empty metric values.
and this case shouldn't affect AVG neither MAX at all.

No sure how looks data where were we found the issue - if they have duplicated records with non-empty metric values.

So it depends how duplication is frequentant (especially for AVGs).

gtanzillo · 2017-10-13T13:12:30Z

I agree that this change should make any calculations based on a range of metrics more accurate. However, it's not apparent what the cause of the dups is and whether they are truly dups - all values are the same or whether they are dups based on timestamp and resource but with different metric values. If it's the latter, is there a risk of throwing away valid metrics?

I guess my question is somewhat beyond the scope of this PR since we were already doing this. And if this make the process more accurate, I'm good with it.

agrare · 2017-10-13T13:22:20Z

@gtanzillo so we believe the cause of the dups is because we have >1 metrics_processor_workers and we have overlapping date ranges queued so they both process the same rollup range at the same time and without unique indexes on the table or some external locking it will be difficult to prevent this. This is something we're planning on addressing in rearch.

@Ladas @jameswnl please correct anything I got wrong in ^^ :D

lpichler · 2017-10-13T13:27:38Z

resource but with different metric values. If it's the latter, is there a risk of throwing away valid metrics?

the method is doing 'merging' values. It means that if any value is missing(is nil) it is populated from other duplicate record with any value. So if duplicated records have values, the winner is the first record with any value. (For chargeback query is ordered by 'resource_id, timestamp' which are same, so I don't know how to determine what is first record in the set of duplicates.)

Remove duplicate metric_rollups not dealing with active relation

agrare · 2017-10-13T13:33:28Z

Nice find @jameswnl !

Ladas · 2017-10-13T14:32:33Z

@gtanzillo yes what @agrare said, we will be adding unique indexes for 5.0, so any duplicates will not be possible. Until then, we need to live with it, since it's pretty hard to avoid duplication.

Remove duplicate metric_rollups not dealing with active relation (cherry picked from commit 15e48b8) https://bugzilla.redhat.com/show_bug.cgi?id=1505417

simaishi · 2017-10-23T15:46:33Z

Euwe backport details:

$ git log -1
commit 76255e3f57ecc6e78e7361b95b430a8602719380
Author: Adam Grare <[email protected]>
Date:   Fri Oct 13 09:32:10 2017 -0400

    Merge pull request #16166 from jameswnl/dedup
    
    Remove duplicate metric_rollups not dealing with active relation
    (cherry picked from commit 15e48b8e96c1aa28e5f91fa67035d14e0c98baf9)
    
    https://bugzilla.redhat.com/show_bug.cgi?id=1505417

Remove duplicate metric_rollups not dealing with active relation (cherry picked from commit 15e48b8) https://bugzilla.redhat.com/show_bug.cgi?id=1505415

simaishi · 2017-11-13T21:35:07Z

Fine backport details:

$ git log -1
commit 5702efb5e85650bbd506b7178138d928ab2e3c2c
Author: Adam Grare <[email protected]>
Date:   Fri Oct 13 09:32:10 2017 -0400

    Merge pull request #16166 from jameswnl/dedup
    
    Remove duplicate metric_rollups not dealing with active relation
    (cherry picked from commit 15e48b8e96c1aa28e5f91fa67035d14e0c98baf9)
    
    https://bugzilla.redhat.com/show_bug.cgi?id=1505415

Remove duplicate metric_rollups not dealing with active relation (cherry picked from commit 15e48b8) https://bugzilla.redhat.com/show_bug.cgi?id=1505415

miq-bot added bug wip labels Oct 10, 2017

lpichler reviewed Oct 11, 2017

View reviewed changes

jameswnl force-pushed the dedup branch 2 times, most recently from cad295a to 9b508f9 Compare October 11, 2017 14:24

remove duplicate metric_rollups not dealing with active relation

d338c6f

jameswnl force-pushed the dedup branch from 9b508f9 to d338c6f Compare October 11, 2017 15:05

jameswnl changed the title ~~[WIP] Remove duplicate metric_rollups not dealing with active relation~~ Remove duplicate metric_rollups not dealing with active relation Oct 11, 2017

miq-bot removed the wip label Oct 11, 2017

lpichler approved these changes Oct 11, 2017

View reviewed changes

Ladas approved these changes Oct 11, 2017

View reviewed changes

chessbyte assigned agrare Oct 12, 2017

miq-bot added euwe/yes labels Oct 12, 2017

agrare reviewed Oct 13, 2017

View reviewed changes

gtanzillo approved these changes Oct 13, 2017

View reviewed changes

agrare approved these changes Oct 13, 2017

View reviewed changes

agrare added this to the Sprint 71 Ending Oct 16, 2017 milestone Oct 13, 2017

agrare added the providers/metrics label Oct 13, 2017

agrare merged commit d338c6f into ManageIQ:master Oct 13, 2017

agrare added a commit that referenced this pull request Oct 13, 2017

Merge pull request #16166 from jameswnl/dedup

15e48b8

Remove duplicate metric_rollups not dealing with active relation

jameswnl deleted the dedup branch October 13, 2017 21:36

simaishi pushed a commit that referenced this pull request Oct 23, 2017

Merge pull request #16166 from jameswnl/dedup

76255e3

Remove duplicate metric_rollups not dealing with active relation (cherry picked from commit 15e48b8) https://bugzilla.redhat.com/show_bug.cgi?id=1505417

simaishi added euwe/backported and removed euwe/yes labels Oct 23, 2017

simaishi pushed a commit that referenced this pull request Nov 13, 2017

Merge pull request #16166 from jameswnl/dedup

5702efb

Remove duplicate metric_rollups not dealing with active relation (cherry picked from commit 15e48b8) https://bugzilla.redhat.com/show_bug.cgi?id=1505415

simaishi added fine/backported and removed fine/yes labels Nov 13, 2017

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Remove duplicate metric_rollups not dealing with active relation #16166

Remove duplicate metric_rollups not dealing with active relation #16166

jameswnl commented Oct 10, 2017

jameswnl commented Oct 10, 2017

jameswnl commented Oct 10, 2017

jameswnl commented Oct 10, 2017

lpichler Oct 11, 2017 •

edited

Loading

jameswnl Oct 11, 2017

miq-bot commented Oct 11, 2017

jameswnl commented Oct 11, 2017

lpichler commented Oct 11, 2017

lpichler left a comment

Ladas left a comment

blomquisg commented Oct 11, 2017

jameswnl commented Oct 12, 2017

agrare Oct 13, 2017

agrare commented Oct 13, 2017

lpichler commented Oct 13, 2017

gtanzillo commented Oct 13, 2017

agrare commented Oct 13, 2017

lpichler commented Oct 13, 2017

agrare commented Oct 13, 2017

Ladas commented Oct 13, 2017

simaishi commented Oct 23, 2017

simaishi commented Nov 13, 2017

Remove duplicate metric_rollups not dealing with active relation #16166

Remove duplicate metric_rollups not dealing with active relation #16166

Conversation

jameswnl commented Oct 10, 2017

jameswnl commented Oct 10, 2017

jameswnl commented Oct 10, 2017

jameswnl commented Oct 10, 2017

lpichler Oct 11, 2017 • edited Loading

Choose a reason for hiding this comment

jameswnl Oct 11, 2017

Choose a reason for hiding this comment

miq-bot commented Oct 11, 2017

jameswnl commented Oct 11, 2017

lpichler commented Oct 11, 2017

lpichler left a comment

Choose a reason for hiding this comment

Ladas left a comment

Choose a reason for hiding this comment

blomquisg commented Oct 11, 2017

jameswnl commented Oct 12, 2017

agrare Oct 13, 2017

Choose a reason for hiding this comment

agrare commented Oct 13, 2017

lpichler commented Oct 13, 2017

gtanzillo commented Oct 13, 2017

agrare commented Oct 13, 2017

lpichler commented Oct 13, 2017

agrare commented Oct 13, 2017

Ladas commented Oct 13, 2017

simaishi commented Oct 23, 2017

simaishi commented Nov 13, 2017

lpichler Oct 11, 2017 •

edited

Loading