Is logti-a3460-17 a valid host name on your system? The sdm is trying to resolve this as an address but it is failing. This resolution needs to work on the nodes as well as the frontend, so the node would need to be configured to resolve the name correctly.
Greg
Thank you. I have applied the patch, but I just get a similar error,
with gethostbyname instead of getaddrinfo.
#PTP job_id=19157
Data for JOB [38102,1] offset 0
SDM: [server] effsize: 2, size: 1, rv: 0
SDM: [server] Found routing file, size=1
SDM: [0] size 2
SDM: [0] in sdm_create_sockd_map
SDM: [0] sdm_route_get_route dest {0}, parent 1
SDM: [0] adjacent nodes: {}
SDM: [0] This node is a server!
SDM: [server] effsize: 2, size: 1, rv: 0
SDM: [0] nodeID: 0, hostname: logti-a3460-17, port: 51969
SDM: [0] found entry for node 0
SDM: [0] found my entry
SDM: [0] [ACCEPT] sockfd: 4
SDM: [0] port bound: 51969
SDM: [0] bound to port 51969
SDM: [master] effsize: 2, size: 1, rv: 0
SDM: [master] Found routing file, size=1
SDM: [1] size 2
SDM: [1] route for 0 is {}
SDM: [1] in sdm_create_sockd_map
SDM: [1] sdm_route_get_route dest {0-1}, parent 1
SDM: [1] adjacent nodes: {0}
SDM: [1] adding 0 to my map
SDM: [master] effsize: 2, size: 1, rv: 0
SDM: [1] nodeID: 0, hostname: logti-a3460-17, port: 51969
gethostbyname: Connection timed out
SDM: [1] gethostbyname error. hostname: logti-a3460-17, port: 51969
SDM: [1] failed to connect to child logti-a3460-17:51969
SDM: [1] sdm_message_init failed
SDM: sdm_init failed
On 14-09-05 05:37 PM, Greg Watson
wrote:
Put the patch file somewhere, the go to the
org.eclipse.ptp.debug.sdm directory and type “patch -p3 <
/path/to/patch_file”. Type “make” to rebuild the sdm.
HTH,
Greg
Yes it is Ubuntu...
Sorry I don't know how to apply the patch, what should I
do?
On 14-09-05 05:12 PM, Greg
Watson wrote:
Valentin,
Is your backend Ubuntu by any chance? I have a
suspicion that getaddrinfo is buggy on Ubuntu, and
this may be causing the problem. If you’re feeling
adventurous, you could try applying the following
patch to the sdm which replaces getaddrinfo with
gethostbyname. It would be interesting to see if
this helps at all.
Regards,
Greg
Hi Greg,
Thanks for your help. I was using Open MPI
1.8.1 and PTP 8.0.1 (with the fix for OMPI
1.8, which was working since the debugger
could launch). I have now updated to PTP
8.1.0 and for some reason I get the
getaddrinfo error again, the same way it was
described by Christoph Schober in a previous
thread (progress bar stuck, cannot connect
to debugger, Eclipse hangs until SDM process
is killed).
This is slightly unrelated to my original
problem but now I cannot start debugging
anymore, and I don't understand why I have
this issue again. I made sure the
start_job.pl script was correct, and I
recompiled SDM from the 8.1.0 source
archive.
Some SDM debug output:
#PTP job_id=15465
Data for JOB [57962,1] offset 0
SDM: [server] effsize: 2, size: 1, rv: 0
SDM: [server] Found routing file, size=1
SDM: [0] size 2
SDM: [0] in sdm_create_sockd_map
SDM: [0] sdm_route_get_route dest {0},
parent 1
SDM: [0] adjacent nodes: {}
SDM: [0] This node is a server!
SDM: [server] effsize: 2, size: 1, rv: 0
SDM: [0] nodeID: 0, hostname:
logti-a3460-17, port: 59885
SDM: [0] found entry for node 0
SDM: [0] found my entry
SDM: [0] [ACCEPT] sockfd: 4
SDM: [0] port bound: 59885
SDM: [0] bound to port 59885
SDM: [master] effsize: 2, size: 1, rv: 0
SDM: [master] Found routing file, size=1
SDM: [1] size 2
SDM: [1] route for 0 is {}
SDM: [1] in sdm_create_sockd_map
SDM: [1] sdm_route_get_route dest {0-1},
parent 1
SDM: [1] adjacent nodes: {0}
SDM: [1] adding 0 to my map
SDM: [master] effsize: 2, size: 1, rv: 0
SDM: [1] nodeID: 0, hostname:
logti-a3460-17, port: 59885
getaddrinfo: Connection timed out
SDM: [1] getaddrinfo error. hostname:
logti-a3460-17, port: 59885
SDM: [1] failed to connect to child
logti-a3460-17:59885
SDM: [1] sdm_message_init failed
SDM: sdm_init failed
On 14-09-03
04:05 PM, Greg Watson wrote:
I just tried
debugging the MPI Pi C++ Project using
Open MPI 1.6 on Ubuntu 14.04 with PTP
8.0.1 and it worked fine.
Can you provide more details on your
setup? Maybe a screenshot showing the
Parallel Debug perspective would help?
Greg
<Mail Attachment.png>
On Sep 3, 2014, at 12:52 PM, Valentin
Vervondel < valentin.vervondel.1@xxxxxxxxxxxxx>
wrote:
Yes I have a
stack frame selected, and it is
displaying some variables just none of
mine, only what seems to be Open MPI
variables, no matter where I break in
my program. Here is a full list:
cl::NullRange (No value)
ompi_stdio_seek_set 0
SEEK_END 2
ompi_stdio_seek_end 2
std::__ioinit {...}
_S_refcount 0
_S_synced_with_stdio false
SEEK_CUR 1
ompi_stdio_seek_cur 1
SEEK_SET 0
On 14-09-03 09:58 AM, Greg Watson
wrote:
Do you have a
stack frame selected in the Debug
view? The Variables view will not
display anything unless you have a
stack frame selected.
Regards,
Greg
On Aug 21, 2014, at 2:52 PM,
Valentin Vervondel <valentin.vervondel.1@xxxxxxxxxxxxx>
wrote:
Hi,
When I start a parallel debug of
my Open MPI application, I don't
have my program variables
displayed in the Variables tab of
the parallel debug perspective. I
actually have two Variables tab,
the first one only contains
variables such as
ompi_stdio_seek_set,
ompi_stdio_seek_end, SEEK_CUR,
cl::NullRange, etc. but none of my
own variables, no matter where I
am debugging in the program. The
second one iw just empty. Same for
when I hover on a variable, no
value is displayed.
I have tried closing the tabs and
resetting the perspective, but the
problem persists. I am using Luna
with PTP 8.0.1 on Ubuntu.
Thanks!
_______________________________________________
ptp-user mailing list
ptp-user@xxxxxxxxxxx
To change your delivery options,
retrieve your password, or
unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/ptp-user
_______________________________________________
ptp-user mailing list
ptp-user@xxxxxxxxxxx
To change your delivery options,
retrieve your password, or
unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/ptp-user
_______________________________________________
ptp-user mailing list
ptp-user@xxxxxxxxxxx
To change your delivery options,
retrieve your password, or unsubscribe
from this list, visit
https://dev.eclipse.org/mailman/listinfo/ptp-user
_______________________________________________
ptp-user mailing list
ptp-user@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/ptp-user
_______________________________________________
ptp-user mailing list
ptp-user@xxxxxxxxxxx
To change your delivery options, retrieve your
password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/ptp-user
_______________________________________________
ptp-user mailing list
ptp-user@xxxxxxxxxxx
To change your delivery options, retrieve your password, or
unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/ptp-user
_______________________________________________
ptp-user mailing list
ptp-user@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/ptp-user
_______________________________________________ ptp-user mailing list ptp-user@xxxxxxxxxxx To change your delivery options, retrieve your password, or unsubscribe from this list, visit https://dev.eclipse.org/mailman/listinfo/ptp-user |