[
https://issues.apache.org/jira/browse/IGNITE-29170?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Dmitry Werner updated IGNITE-29170:
-----------------------------------
Description:
Currently, long running cache futures are only reported via the log warning
"First %d long running cache futures [total=%d]" from
GridCachePartitionExchangeManager.
There is no way to monitor them without parsing logs.
It is proposed to add two metrics to the tx metrics registry
(TransactionMetricsAdapter):
- LongRunningCacheFuturesCount — the number of in-progress cache futures whose
duration exceeds longOperationsDumpTimeout;
- MaxCacheFutureDuration — the maximum duration of in-progress cache futures,
in milliseconds.
Both are computed from GridCacheMvccManager active and atomic futures, so they
are exposed through all metric subsystems (JMX, exporters, system views) with
no log parsing required.
The existing warning behavior is unchanged. Covered by
GridCachePartitionExchangeManagerWarningsTest; documented in
monitoring-metrics/new-metrics.adoc.
was:
Currently, the event "First %d long running cache futures [total=%d]" is only
reported via the log by GridCachePartitionExchangeManager.
There is no way to track how often such events occur without parsing (grepping)
the log.
We need to expose this as a metric so the count of events can be monitored
through the metrics subsystem (e.g. JMX, system views, exporters) instead of
relying on log parsing.
> Add a metric for long running cache futures
> -------------------------------------------
>
> Key: IGNITE-29170
> URL: https://issues.apache.org/jira/browse/IGNITE-29170
> Project: Ignite
> Issue Type: Task
> Reporter: Dmitry Werner
> Assignee: Dmitry Werner
> Priority: Major
> Labels: metrics-to-log
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> Currently, long running cache futures are only reported via the log warning
> "First %d long running cache futures [total=%d]" from
> GridCachePartitionExchangeManager.
> There is no way to monitor them without parsing logs.
> It is proposed to add two metrics to the tx metrics registry
> (TransactionMetricsAdapter):
> - LongRunningCacheFuturesCount — the number of in-progress cache futures
> whose duration exceeds longOperationsDumpTimeout;
> - MaxCacheFutureDuration — the maximum duration of in-progress cache futures,
> in milliseconds.
> Both are computed from GridCacheMvccManager active and atomic futures, so
> they are exposed through all metric subsystems (JMX, exporters, system views)
> with no log parsing required.
> The existing warning behavior is unchanged. Covered by
> GridCachePartitionExchangeManagerWarningsTest; documented in
> monitoring-metrics/new-metrics.adoc.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
