You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@ignite.apache.org by Вадим Опольский <va...@gmail.com> on 2017/06/04 10:43:19 UTC

Fwd: place for printing info about partition distribution (IGNITE-4756)

Guys, lets suggest the format for this log.

Yakov's version is "print only nodes that differ on +/-10% from average
partitions count we can print problem message looks like:

[cacheName=NAME, nodes=[nodeId=<ID, totalPartitionsCount=<count>,
percentageOfTotalPartsCount=<percentage>, parts=[<count>, ...,<count>]]

Where the first element of parts array is a count of primary partitions,
the second is count of backup0 partitions etc.
 The default value of the property IGNITE_PART_DISTRIBUTION_WARN_THRESHOLD
must disable distribution calculation because it takes <caches_count> *
<parttitons_count> * <nodes_count> operations per node. "

Do you agree ?


---------- Forwarded message ----------
From: Вадим Опольский <va...@gmail.com>
Date: 2017-06-04 13:42 GMT+03:00
Subject: Fwd: place for printing info about partition distribution
(IGNITE-4756)
To: Denis Magda <dm...@apache.org>, tledkov@gridgain.com, Yakov Zhdanov <
yzhdanov@gridgain.com>


Hello guys!

I cant start resolve issue 4756 (https://issues.apache.org/
jira/browse/IGNITE-4756) because don't understand place for pinting info
about partition.

Yakov, Taras do you have any idea about it ?

Vadim Opolski



---------- Forwarded message ----------
From: Вадим Опольский <va...@gmail.com>
Date: 2017-05-31 18:24 GMT+03:00
Subject: place for printing info about partition distribution
To: dev@ignite.apache.org, Denis Magda <dm...@apache.org>,
tledkov@gridgain.com, Yakov Zhdanov <yz...@gridgain.com>


Hello guys!

I want to discuss and suggest some format and place for pinting info about
partition distribution to log from issue https://issues.apache.org/jira
/browse/IGNITE-4756

What do you think about it ?

Issue is about RendezvousAffinityFunction and FairAffinityFunction classes.

Vadim

Re: place for printing info about partition distribution (IGNITE-4756)

Posted by Taras Ledkov <tl...@gridgain.com>.
According to the last Yakov's comments at the ticket:

1. Calculation and printing distribution deviation will be produced only 
for local node.
2. In this case it takes only one additional compare nodeID if we build 
statistic with the assignment calculation.
3. We have return injection of the Ignite instance into the AF to gather 
local node.


On 05.06.2017 20:39, Denis Magda wrote:
> Taras, Yakov,
>
> Please join the conversation.
>
> —
> Denis
>
>> On Jun 4, 2017, at 3:43 AM, Вадим Опольский <vaopolskij@gmail.com 
>> <ma...@gmail.com>> wrote:
>>
>> Guys, lets suggest the format for this log.
>>
>> Yakov's version is "print only nodes that differ on +/-10% from 
>> average partitions count we can print problem message looks like:
>>
>> [cacheName=NAME, nodes=[nodeId=<ID, totalPartitionsCount=<count>, 
>> percentageOfTotalPartsCount=<percentage>, parts=[<count>, ...,<count>]]
>>
>> Where the first element of parts array is a count of primary 
>> partitions, the second is count of backup0 partitions etc.
>>  The default value of the property 
>> IGNITE_PART_DISTRIBUTION_WARN_THRESHOLD must disable distribution 
>> calculation because it takes <caches_count> * <parttitons_count> * 
>> <nodes_count> operations per node. "
>>
>> Do you agree ?
>>
>>
>> ---------- Forwarded message ----------
>> From: *Вадим Опольский* <vaopolskij@gmail.com 
>> <ma...@gmail.com>>
>> Date: 2017-06-04 13:42 GMT+03:00
>> Subject: Fwd: place for printing info about partition distribution 
>> (IGNITE-4756)
>> To: Denis Magda <dmagda@apache.org <ma...@apache.org>>, 
>> tledkov@gridgain.com <ma...@gridgain.com>, Yakov Zhdanov 
>> <yzhdanov@gridgain.com <ma...@gridgain.com>>
>>
>>
>> Hello guys!
>>
>> I cant start resolve issue 4756 
>> (https://issues.apache.org/jira/browse/IGNITE-4756 
>> <https://issues.apache.org/jira/browse/IGNITE-4756>) because don't 
>> understand place for pinting info about partition.
>>
>> Yakov, Taras do you have any idea about it ?
>>
>> Vadim Opolski
>>
>>
>>
>> ---------- Forwarded message ----------
>> From: *Вадим Опольский* <vaopolskij@gmail.com 
>> <ma...@gmail.com>>
>> Date: 2017-05-31 18:24 GMT+03:00
>> Subject: place for printing info about partition distribution
>> To: dev@ignite.apache.org <ma...@ignite.apache.org>, Denis Magda 
>> <dmagda@apache.org <ma...@apache.org>>, tledkov@gridgain.com 
>> <ma...@gridgain.com>, Yakov Zhdanov <yzhdanov@gridgain.com 
>> <ma...@gridgain.com>>
>>
>>
>> Hello guys!
>>
>> I want to discuss and suggest some format and place for pinting info 
>> about partition distribution to log from issue 
>> https://issues.apache.org/jira/browse/IGNITE-4756 
>> <https://issues.apache.org/jira/browse/IGNITE-4756>
>>
>> What do you think about it ?
>>
>> Issue is about RendezvousAffinityFunction and FairAffinityFunction 
>> classes.
>>
>> Vadim
>>
>>
>>
>

-- 
Taras Ledkov
Mail-To: tledkov@gridgain.com


Re: place for printing info about partition distribution (IGNITE-4756)

Posted by Denis Magda <dm...@apache.org>.
Taras, Yakov,

Please join the conversation.

—
Denis

> On Jun 4, 2017, at 3:43 AM, Вадим Опольский <va...@gmail.com> wrote:
> 
> Guys, lets suggest the format for this log.
> 
> Yakov's version is "print only nodes that differ on +/-10% from average partitions count we can print problem message looks like:
> 
> [cacheName=NAME, nodes=[nodeId=<ID, totalPartitionsCount=<count>, percentageOfTotalPartsCount=<percentage>, parts=[<count>, ...,<count>]]
> 
> Where the first element of parts array is a count of primary partitions, the second is count of backup0 partitions etc.
>  The default value of the property IGNITE_PART_DISTRIBUTION_WARN_THRESHOLD must disable distribution calculation because it takes <caches_count> * <parttitons_count> * <nodes_count> operations per node. "
> 
> Do you agree ?
> 
> 
> ---------- Forwarded message ----------
> From: Вадим Опольский <vaopolskij@gmail.com <ma...@gmail.com>>
> Date: 2017-06-04 13:42 GMT+03:00
> Subject: Fwd: place for printing info about partition distribution (IGNITE-4756)
> To: Denis Magda <dmagda@apache.org <ma...@apache.org>>, tledkov@gridgain.com <ma...@gridgain.com>, Yakov Zhdanov <yzhdanov@gridgain.com <ma...@gridgain.com>>
> 
> 
> Hello guys!
> 
> I cant start resolve issue 4756 (https://issues.apache.org/jira/browse/IGNITE-4756 <https://issues.apache.org/jira/browse/IGNITE-4756>) because don't understand place for pinting info about partition.
> 
> Yakov, Taras do you have any idea about it ?
> 
> Vadim Opolski
> 
> 
> 
> ---------- Forwarded message ----------
> From: Вадим Опольский <vaopolskij@gmail.com <ma...@gmail.com>>
> Date: 2017-05-31 18:24 GMT+03:00
> Subject: place for printing info about partition distribution
> To: dev@ignite.apache.org <ma...@ignite.apache.org>, Denis Magda <dmagda@apache.org <ma...@apache.org>>, tledkov@gridgain.com <ma...@gridgain.com>, Yakov Zhdanov <yzhdanov@gridgain.com <ma...@gridgain.com>>
> 
> 
> Hello guys!
> 
> I want to discuss and suggest some format and place for pinting info about partition distribution to log from issue https://issues.apache.org/jira/browse/IGNITE-4756 <https://issues.apache.org/jira/browse/IGNITE-4756>
> What do you think about it ?
> 
> Issue is about RendezvousAffinityFunction and FairAffinityFunction classes.
> 
> Vadim
> 
> 
>