RE: Data region LRU offheap algo not working

[rick_tem]

Hi,

I have created the following JIRA for this issue.

IGNITE-12215

Thanks,
Rick



--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/

RE: Data region LRU offheap algo not working

[rick_tem]

Hi, 

Please find enclosed the memory crash with the following configuration along
with the data region metrics just before the crash.  This should be easy to
reproduce.  Simply keep injecting data into the cache at a high rate with a
small data region size.

Thanks,
Rick












 


Ignite-dataregion.png

  
Ignite.log
  



--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/

RE: Data region LRU offheap algo not working

[rick_tem]

Thanks for your reply Alexandr.

However, looking at the log, memory continues to be lost.  You continue to
see loss of memory after this message:
2019-09-09 11:04:03.557 WARN  [sys-stripe-5-#6%TemenosGrid%]
IgniteCacheDatabaseSharedManager - Page-based evictions started. Consider
increasing 'maxSize' on Data Region configuration: 1G_Region

Log is: 
http://apache-ignite-users.70518.x6.nabble.com/file/t869/dateRepo1.out

While not in this log, this eventually causes an out of memory issue stack
trace that is unrecoverable.  I'll get the log that shows that and turn on
the metrics to get better information.  It's pretty easy for me to
reproduce.

Thanks,
Rick








--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/

RE: Data region LRU offheap algo not working

[Alexandr Shapkin]

Hello!

As far as my understanding goes, I see that eviction works ok.

>2019-09-09 11:04:03.557 WARN  [sys-stripe-5-#6%TemenosGrid%] 
>IgniteCacheDatabaseSharedManager - Page-based evictions started. Consider 
>increasing 'maxSize' on Data Region configuration: 1G_Region

This message tells that now eviction process is started and will free pages 
according to the algorithm.
it means that when Ignite requests for a new page and there are no available 
ones, then Ignite will apply eviction first and then use those pages for a new 
values.
So you will not see any significant memory changes during eviction. 
I.e. Ignite will show that the data region is full, but still allows you to 
write new values.

There are no any special messages log messages regarding evicted pages. Only 
the first one, when eviction comes to the play.
I think you may try to look into getEvictionRate() metrics [1] for details. 
Also you should be able to see it in Visor GUI.

[1] - 
https://apacheignite.readme.io/docs/memory-metrics#section-data-region-metrics

From: rick_tem
Sent: Tuesday, September 10, 2019 10:12 AM
To: user@ignite.apache.org
Subject: Data region LRU offheap algo not working

Hi,

I am trying to find out why it appears the RANDOM_LRU algo doesn't seem to
work with the following config.  Log attached, as well...after the log of
the below...

2019-09-09 11:04:03.557 WARN  [sys-stripe-5-#6%TemenosGrid%]
IgniteCacheDatabaseSharedManager - Page-based evictions started. Consider
increasing 'maxSize' on Data Region configuration: 1G_Region

after a few minutes you see memory steadily decrease.  What information in
the log will help me determine how many pages are freed, etc?

Thanks,
Rick

dateRepo1.out
  
dataRepo2.out
  

























--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/

Re: Data region LRU offheap algo not working

[rick_tem]

Hi, it's in the log 2.7.5...Thanks for any help here!
Thanks,
Rick



--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/

Re: Data region LRU offheap algo not working

[Michaelikus]

what is tour ignite version?



-
Russian telegram Ignite chat at https://t.me/RU_Ignite
--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/