[ 
https://issues.apache.org/jira/browse/IGNITE-29277?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Nikita Amelchev updated IGNITE-29277:
-------------------------------------
    Description: {{MaxExplicitLockHoldTime}} reports only the longest hold time 
among the explicit locks currently held, and {{ExplicitLockHoldTimeHistogram}} 
accounts for a lock only after it is released, which its name does not say. So 
there is no way to see how many locks are being held for too long right now: a 
single forgotten lock cannot be told apart from a mass leak, and an alert 
cannot be set on the number of such locks. A histogram of hold times of the 
currently held locks is needed, with the two histograms named as a pair, 
{{CurrentExplicitLockHoldTime}} and {{{}ReleasedExplicitLockHoldTime{}}}, 
before the existing one is first released.  (was: {{MaxExplicitLockHoldTime}} 
reports only the longest hold time among the explicit locks currently held, and 
{{ExplicitLockHoldTimeHistogram}} accounts for a lock only after it is 
released. So there is no way to see how many locks are being held for too long 
right now: a single forgotten lock cannot be told apart from a mass leak, and 
an alert cannot be set on the number of such locks. A histogram of hold times 
of the currently held locks would show it, with the threshold of a long-held 
lock configured as the histogram bound.)

> Add a histogram of currently held explicit locks
> ------------------------------------------------
>
>                 Key: IGNITE-29277
>                 URL: https://issues.apache.org/jira/browse/IGNITE-29277
>             Project: Ignite
>          Issue Type: Task
>            Reporter: Nikita Amelchev
>            Assignee: Nikita Amelchev
>            Priority: Major
>             Fix For: 2.20
>
>          Time Spent: 10m
>  Remaining Estimate: 0h
>
> {{MaxExplicitLockHoldTime}} reports only the longest hold time among the 
> explicit locks currently held, and {{ExplicitLockHoldTimeHistogram}} accounts 
> for a lock only after it is released, which its name does not say. So there 
> is no way to see how many locks are being held for too long right now: a 
> single forgotten lock cannot be told apart from a mass leak, and an alert 
> cannot be set on the number of such locks. A histogram of hold times of the 
> currently held locks is needed, with the two histograms named as a pair, 
> {{CurrentExplicitLockHoldTime}} and {{{}ReleasedExplicitLockHoldTime{}}}, 
> before the existing one is first released.



--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

Reply via email to