dmitrygx

@krakowski I tried to reproduce the issue using you reproducer and setting `UCX_TLS=tcp`. But it works fine with 1.12.1 and master.

> It works for me most of the time, too. how many iterations does it take to catch this? > I noticed that the connection gets established as soon as...

@krakowski thanks! but it still works fine using this dc9243556ec10178fc9fc2f8be5e17f2bf79cd19 commit. see the output of the client (it is mixed due to threads contention): ``` $UCX_TLS=tcp ./ucp_am_mt client >> Initializing...

> I also could not really find out what `max_iov` stands for. Can you please explain briefly what the assertion checks? it expects that transport's supported number of IOVs will...

@krakowski thansk for collecting inforamtion and helping me with the reproduction! I still can't reproduce the issue locally, but I understand what's wrong now. Actually, it seems that it is...

@krakowski thanks for testing it! I'll think about upstreaming it. > I noticed that the first transport lane after reconfiguration changed from posix/memory to sysv/memory. yeah, interesting... I compared locally...

@krakowski thanks you! It is very helpful. yes, v1.12.1 has a difference with selection between sysv and posix in compare to master and my local branch. It seems that something...

@hoopoepg @evgeny-leksikov could you review pls?

@brminich could you review pls if have time?

what is the UCX version used by you? I think that the `neth->psn < ep->rx.ooo_pkts.head_sn` (i.e. `1` < `3`) case should be handled here: https://github.com/openucx/ucx/blob/488cb507f12dbb6904bd7e7058bb1e7206ddc525/src/uct/ib/ud/base/ud_ep.c#L815 see https://github.com/openucx/ucx/pull/7353 PR description which...