"Failed to update store" cacheCriticalSize

[Mike Wiesenberg <mi...@gmail.com>]

Hi,
 Using GridCacheWriteBehindStore (Version 2.10.0), we are occasionally
receiving this error message

"Failed to update store (value will be lost as current buffer size is
greater than 'cacheCriticalSize.."

Looking at the code to try to determine how to raise the buffer size, I
observe that

1. This only occurs when there is an exception writing to the store
("Unable to update"). I observe 18 such failures in my log but 512 "Failed
to update messages", which indicates that the sql errors are negatively
impacting other values which happen to have been in the same batch with
store errors. Is there a way to ignore or skip the 'Unable to update" fails
so other values are not lost when those occur? (It seems kind of random
that we only discard values due to the cacheCriticalSize when an unrelated
exception occurs. Why?)

2. I was looking to see if I can increase the cacheCriticalSize, but it's
calculated by multiplying the cacheMaxSize times 1.5, and increasing the
cacheMaxSize would also increase the number of values held before storing,
so it wouldn't help much. Is there another way to adjust configuration to
mitigiate this problem?


Thanks,
 Mike

Re: "Failed to update store" cacheCriticalSize

[Ilya Kazakov <ka...@gmail.com>]

Hi Mike. In your case try one of two ways.

In cache configuration:
1. Increase writeBehindFlushSize parameter (default is 10240 bytes).
or
2. Decrease writeBehindFlushFreq parameter (default is 5000 ms).

https://ignite.apache.org/releases/latest/javadoc/org/apache/ignite/configuration/CacheConfiguration.html

Hope this will help you.

----------------------
Ilya

пт, 12 нояб. 2021 г. в 03:29, Mike Wiesenberg <mi...@gmail.com>:

> Hi,
>  Using GridCacheWriteBehindStore (Version 2.10.0), we are occasionally
> receiving this error message
>
> "Failed to update store (value will be lost as current buffer size is
> greater than 'cacheCriticalSize.."
>
> Looking at the code to try to determine how to raise the buffer size, I
> observe that
>
> 1. This only occurs when there is an exception writing to the store
> ("Unable to update"). I observe 18 such failures in my log but 512 "Failed
> to update messages", which indicates that the sql errors are negatively
> impacting other values which happen to have been in the same batch with
> store errors. Is there a way to ignore or skip the 'Unable to update" fails
> so other values are not lost when those occur? (It seems kind of random
> that we only discard values due to the cacheCriticalSize when an unrelated
> exception occurs. Why?)
>
> 2. I was looking to see if I can increase the cacheCriticalSize, but it's
> calculated by multiplying the cacheMaxSize times 1.5, and increasing the
> cacheMaxSize would also increase the number of values held before storing,
> so it wouldn't help much. Is there another way to adjust configuration to
> mitigiate this problem?
>
>
> Thanks,
>  Mike
>