You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
mattab opened this issue
Nov 13, 2014
· 0 comments
Assignees
Labels
BugFor errors / faults / flaws / inconsistencies etc.CriticalIndicates the severity of an issue is very critical and the issue has a very high priority.
Set as Critical priority as it's blocking some work for Enterprise client. there are data discrepencies between the Live instance and the instance built with replayed logs.
Explanation
in the log importer we do not properly put all the requests for a given visitor into the same bulk requests, and maybe requests by same visitor ends up being inserted at the same time in two or more bulk requests. The logic for creating queues is here: https://github.com/piwik/piwik/blob/master/misc/log-analytics/import_logs.py#L1161-L1166 - it may be fragile.
This is causing an issue when
trust_visitors_cookies = 1 in the config.
--replay-tracking is used in the log importer
Proposed solution
Currently requests are grouped by IP address.
New proposed logic:
if uid is set, use it,
else if cid is set, use it,
else if _id is set, use it,
else if cip is set, use it,
else use IP address.
This should then solve the problem with trust_visitors_cookies = 1 as in all cases (even when the setting is 0) the requests will be grouped by Visitor ID or Visitor IP (with priority to Visitor ID).
The text was updated successfully, but these errors were encountered:
mattab
added
Bug
For errors / faults / flaws / inconsistencies etc.
Critical
Indicates the severity of an issue is very critical and the issue has a very high priority.
labels
Nov 13, 2014
BugFor errors / faults / flaws / inconsistencies etc.CriticalIndicates the severity of an issue is very critical and the issue has a very high priority.
Set as Critical priority as it's blocking some work for Enterprise client. there are data discrepencies between the Live instance and the instance built with replayed logs.
Explanation
in the log importer we do not properly put all the requests for a given visitor into the same bulk requests, and maybe requests by same visitor ends up being inserted at the same time in two or more bulk requests. The logic for creating queues is here: https://github.com/piwik/piwik/blob/master/misc/log-analytics/import_logs.py#L1161-L1166 - it may be fragile.
This is causing an issue when
trust_visitors_cookies = 1
in the config.--replay-tracking
is used in the log importerProposed solution
Currently requests are grouped by IP address.
New proposed logic:
uid
is set, use it,cid
is set, use it,_id
is set, use it,cip
is set, use it,This should then solve the problem with
trust_visitors_cookies = 1
as in all cases (even when the setting is0
) the requests will be grouped by Visitor ID or Visitor IP (with priority to Visitor ID).The text was updated successfully, but these errors were encountered: