It’s not really about the volume of transactions, more how likely it is that
multiple clients will be writing to the same records.
https://ignite.apache.org/docs/latest/key-value-api/transactions
> On 3 Nov 2021, at 12:26, Sumit Deshinge wrote:
>
> Thank you Stephen. I will perform the next se
Thank you Stephen. I will perform the next set of runs by adding both
TxTimeoutOnPartitionMapExchange and defaultTxTimeout.
*Also, what is the recommendation for transaction concurrency and isolation
settings for very high frequency transactions with a large number of thin
client connections ?*
Yes. And defaultTxTimeout for similar reasons.
Frankly, these are terrible defaults. There is a ticket to fix the latter:
https://issues.apache.org/jira/browse/IGNITE-13064
> On 3 Nov 2021, at 11:59, Sumit Deshinge wrote:
>
> Hi Ilya,
>
> I have tried to run this with both ignite 2.10 and 2.1
Hi Ilya,
I have tried to run this with both ignite 2.10 and 2.11
There are 24 thin clients on different VMs which are trying to perform
*CRUD* operations on the *same AND/OR different caches* parallelly.
If I set the *TxTimeoutOnPartitionMapExchange *property to some value, then
the transactions s
Hi Sumit,
What is Ignite version that you use?
AFAIK partition map exchange is a king of "stop the world" actiity for
the cluster, so any other actions with cluster (like cache creation)
would be suspended until PME end.
If all of your clients concurrently try to create same cache it's OK
that