Ignite Client Hangs forever while connecting when already other Clients Joined the Cluster for a long time

[kotamrajuyashasvi <ko...@gmail.com>]

Hi

There are some Ignite Clients connected to a Cluster, performing
transactional operations.
Then after some time an Ignite Client tries to connect to the cluster . But
the Client hangs up 
during connection.

In the Ignite server logs we could find the following:
12:21:45,268][WARNING]exchange-worker-#118%null%[diagnostic] >>>
GridDhtPartitionsExchangeFuture [topVer=AffinityTopologyVersion [topVer=163,
minorTopVer=0], evt=10, evtNode=TcpDiscoveryNode
[id=2bd97b73-5aac-492e-b28f-2c12ec89540d, addrs=[10.144.198.243,
10.144.44.233, 127.0.0.1, 192.168.144.233], sockAddrs=[10.155.44.10:0,
/127.0.0.1:0, /192.168.12.212:0, /10.144.198.243:0], discPort=0, order=163,
intOrder=84, lastExchangeTime=1521122324954, loc=false, ver=2.1.0#,
isClient=true], done=false]
[12:21:45,268][WARNING]exchange-worker-#118%null%[diagnostic] Pending
transactions:

on hung up client we can see:
[10:51:53,527][WARNING][main][GridCachePartitionExchangeManager] Failed to
wait for initial partition map exchange. Possible reasons are: 
^-- Transactions in deadlock.
^-- Long running transactions (ignore if this is the case).
^-- Unreleased explicit locks.
[10:53:13,541][WARNING][main][GridCachePartitionExchangeManager] Still
waiting for initial partition map exchange .. and this message keeps on
continuing..

The new Client did not even start a Transaction. Why is it getting hung up
during connection? If it is trying to lock a key which is in deadlock then
it can hang up till the transaction timeout, but that's not what is
happening here. Even the Clients already connected are not transaction
timing out(which has been set for 60 secs) which would happen if there was a
deadlock and they are continuously running. What would be the reason the
Client is hanging up forever. Are there any configurations that needs so be
done additionally?







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

Re: Ignite Client Hangs forever while connecting when already other Clients Joined the Cluster for a long time

[Andrey Mashenkov <an...@gmail.com>]

Hi,

Usually, there should be smth that leads to "Failed to wait for initial
partition map exchange.".
It may be an Exception on server node that makes transaction hanged or
prevent sending topology update to client.

Partition map exchange is always wait for all transactions finished and for
locks are released.
And vice versa, transactions\locks are always waiting for exchange has
finished.

What version of Ignite do you use? Have you tried the latest one?
Would you please check if there were no Exceptions on both, client and
server sides.


On Tue, Mar 20, 2018 at 5:09 PM, kotamrajuyashasvi <
kotamrajuyashasvi@gmail.com> wrote:

> Hi
>
> There are some Ignite Clients connected to a Cluster, performing
> transactional operations.
> Then after some time an Ignite Client tries to connect to the cluster . But
> the Client hangs up
> during connection.
>
> In the Ignite server logs we could find the following:
> 12:21:45,268][WARNING]exchange-worker-#118%null%[diagnostic] >>>
> GridDhtPartitionsExchangeFuture [topVer=AffinityTopologyVersion
> [topVer=163,
> minorTopVer=0], evt=10, evtNode=TcpDiscoveryNode
> [id=2bd97b73-5aac-492e-b28f-2c12ec89540d, addrs=[10.144.198.243,
> 10.144.44.233, 127.0.0.1, 192.168.144.233], sockAddrs=[10.155.44.10:0,
> /127.0.0.1:0, /192.168.12.212:0, /10.144.198.243:0], discPort=0,
> order=163,
> intOrder=84, lastExchangeTime=1521122324954, loc=false, ver=2.1.0#,
> isClient=true], done=false]
> [12:21:45,268][WARNING]exchange-worker-#118%null%[diagnostic] Pending
> transactions:
>
> on hung up client we can see:
> [10:51:53,527][WARNING][main][GridCachePartitionExchangeManager] Failed to
> wait for initial partition map exchange. Possible reasons are:
> ^-- Transactions in deadlock.
> ^-- Long running transactions (ignore if this is the case).
> ^-- Unreleased explicit locks.
> [10:53:13,541][WARNING][main][GridCachePartitionExchangeManager] Still
> waiting for initial partition map exchange .. and this message keeps on
> continuing..
>
> The new Client did not even start a Transaction. Why is it getting hung up
> during connection? If it is trying to lock a key which is in deadlock then
> it can hang up till the transaction timeout, but that's not what is
> happening here. Even the Clients already connected are not transaction
> timing out(which has been set for 60 secs) which would happen if there was
> a
> deadlock and they are continuously running. What would be the reason the
> Client is hanging up forever. Are there any configurations that needs so be
> done additionally?
>
>
>
>
>
>
>
> --
> Sent from: http://apache-ignite-users.70518.x6.nabble.com/
>



-- 
Best regards,
Andrey V. Mashenkov