This is reported upstream (UCX) here:
https://github.com/openucx/ucx/issues/5443
On 23/07/2020 06:33, Drew Parsons wrote:
Package: libopenmpi3
Version: 4.0.4-2
Followup-For: Bug #965352
Control: affects -1 src:scalapack
UCX seems to be affecting the scalapack build also:
87/96 Test #82: xdgsep ........................... Passed 109.38 sec
Start 95: xshseqr
88/96 Test #83: xcgsep ........................... Passed 101.14 sec
Start 96: xdhseqr
89/96 Test #96: xdhseqr ..........................***Failed 49.20 sec
ScaLAPACK Test for PDHSEQR
epsilon = 1.1102230246251565E-016
threshold = 30.000000000000000
Residual and Orthogonality Residual computed by:
Residual = || T - Q^T*A*Q ||_F / ( ||A||_F * eps * sqrt(N) )
Orthogonality = MAX( || I - Q^T*Q ||_F, || I - Q*Q^T ||_F ) / (eps * N)
Test passes if both residuals are less then threshold
N NB P Q QR Time CHECK
----- --- ---- ---- -------- ------
[1595480623.088652] [monte:1320201:0] sock.c:344 UCX ERROR
recv(fd=28) failed: Bad address
[1595480623.199533] [monte:1320189:0] sock.c:344 UCX ERROR
sendv(fd=30) failed: Connection reset by peer
[monte:1320189] *** An error occurred in MPI_Bcast
[monte:1320189] *** reported by process [1297350657,2]
[monte:1320189] *** on communicator MPI COMMUNICATOR 5 SPLIT FROM 3
[monte:1320189] *** MPI_ERR_OTHER: known error not in list
[monte:1320189] *** MPI_ERRORS_ARE_FATAL (processes in this communicator will
now abort,
[monte:1320189] *** and potentially your MPI job)
--
Alastair McKinstry, email: alast...@sceal.ie, matrix: @alastair:sceal.ie,
phone: 087-6847928
Green Party Councillor, Galway County Council
