here''s an odd one. Had a client system reboot, and when it came back ypbind was listening on port 988 and a lustre start complained bitterly: Lustre: OBD class driver, http://www.lustre.org/ Lustre: Lustre Version: 1.6.7 Lustre: Build Version: 1.6.7-19691231170000- PRISTINE-.cache.build.BUILD.lustre-kernel-2.6.18.lustre.linu x-2.6.18-92.1.17.el5_lustre.1.6.7smp Lustre: Added LNI 192.5.200.113 at tcp [8/256] LustreError: 6270:0:(linux-tcpip.c:569:libcfs_sock_listen()) Can''t create socket: port 988 already in use LustreError: 122-1: Can''t start acceptor on port 988: port already in use Lustre: Removed LNI 192.5.200.113 at tcp LustreError: 6191:0:(events.c:713:ptlrpc_init_portals()) network initialisation failed # netstat -anp|grep 988 tcp 0 0 0.0.0.0:988 0.0.0.0:* LISTEN 4199/ypbind # netstat -anp|grep ypb tcp 0 0 0.0.0.0:988 0.0.0.0:* LISTEN 4199/ypbind udp 0 0 0.0.0.0:705 0.0.0.0:* 4199/ypbind udp 0 0 0.0.0.0:985 0.0.0.0:* 4199/ypbind Has anyone run into this? I didn''t see mention of it in the mailing list archives or in bugzilla. Lustre 1.6.7, Centos 5.2 on x86_64. Thanks, -bob
On Tue, 2009-03-31 at 17:00 -0500, Robert Olson wrote:> here''s an odd one. Had a client system reboot, and when it came back > ypbind was listening on port 988 and a lustre start complained bitterly:...> Has anyone run into this?Yes.> I didn''t see mention of it in the mailing > list archives or in bugzilla.There is a most certainly a long and winding bug about it. Was filed and close many years ago. It''s just one of those problems can can''t be fixed unless other software (i.e. the rpc portmapper) wants to accept a fix, which I don''t think it does. Ideally one should be able to blacklist ports from the portmapper, but one cannot. b. -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 197 bytes Desc: This is a digitally signed message part Url : http://lists.lustre.org/pipermail/lustre-discuss/attachments/20090331/69b790f5/attachment.bin
Brian J. Murrell wrote:> On Tue, 2009-03-31 at 17:00 -0500, Robert Olson wrote: >> here''s an odd one. Had a client system reboot, and when it came back >> ypbind was listening on port 988 and a lustre start complained bitterly: > > ... > >> Has anyone run into this? > > Yes. >You can work around it with the portreserve package. http://cyberelk.net/tim/software/portreserve/ Cheers, Guy -- Dr. Guy Coates, Informatics System Group The Wellcome Trust Sanger Institute, Hinxton, Cambridge, CB10 1HH, UK Tel: +44 (0)1223 834244 x 6925 Fax: +44 (0)1223 496802 -- The Wellcome Trust Sanger Institute is operated by Genome Research Limited, a charity registered in England with number 1021457 and a company registered in England with number 2742969, whose registered office is 215 Euston Road, London, NW1 2BE.
Guy Coates schrieb:> Brian J. Murrell wrote: >> On Tue, 2009-03-31 at 17:00 -0500, Robert Olson wrote: >>> here''s an odd one. Had a client system reboot, and when it came back >>> ypbind was listening on port 988 and a lustre start complained bitterly: >> ... >> >>> Has anyone run into this? >> Yes. >> > You can work around it with the portreserve package. > > http://cyberelk.net/tim/software/portreserve/Simply ensure the lustre modules are loaded as early as possible. We use the following service script for that purpose. #!/bin/sh # # /etc/rc.lustre-reserveport # # Load Lustre module ptlrpc early in boot process. This is a hack to prevent # portmapper from handing out port 988 to different applications before # Lustre''s network layer has a chance to grab it. case "$1" in start|restart|reload) modprobe ptlrpc || : # XXX The above apparently prevents auto-loading of module # lustre during mount. Force it in now. modprobe lustre || : ;; stop) if which lustre_rmmod >/dev/null; then lustre_rmmod fi ;; esac exit 0 Regards, Uli -- Vorstand/Board of Management: Dr. Bernd Finkbeiner, Dr. Roland Niemeier, Dr. Arno Steitz, Dr. Ingrid Zech Vorsitzender des Aufsichtsrats/ Chairman of the Supervisory Board: Michel Lepert Sitz/Registered Office: Tuebingen Registergericht/Registration Court: Stuttgart Registernummer/Commercial Register No.: HRB 382196
On Apr 1, 2009, at 10:04 AM, Ulrich Sibiller wrote:> Guy Coates schrieb: >> Brian J. Murrell wrote: >>> On Tue, 2009-03-31 at 17:00 -0500, Robert Olson wrote: >>>> here''s an odd one. Had a client system reboot, and when it came >>>> back >>>> ypbind was listening on port 988 and a lustre start complained >>>> bitterly: >>> ... >>> >>>> Has anyone run into this? >>> Yes. >>> >> You can work around it with the portreserve package. >> >> http://cyberelk.net/tim/software/portreserve/ > > Simply ensure the lustre modules are loaded as early as possible. We > use the following service > script for that purpose.Thank you to all for your suggestions. In particular the one above jogs my memory - I had done exactly that in a previous test incarnation of a lustre service here by hacking a "modprobe lustre" into the ypbind startup script. Having an actual startup script to do it is just a little cleaner :-). --bob
On Wed, 2009-04-01 at 10:38 -0500, Robert Olson wrote:> > Thank you to all for your suggestions. In particular the one above > jogs my memory - I had done exactly that in a previous test > incarnation of a lustre service here by hacking a "modprobe lustre" > into the ypbind startup script. Having an actual startup script to do > it is just a little cleaner :-).Well, if you are going to take that approach, don''t most (all?) distros provide a hook to manually load modules. RHEL5 does. I''d think simply adding lnet there would take care of it rather than writing whole initscripts to do it. b. -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 197 bytes Desc: This is a digitally signed message part Url : http://lists.lustre.org/pipermail/lustre-discuss/attachments/20090401/7467da0d/attachment.bin
oh right, of course. lnet is sufficient for triggering what is needed? On Apr 1, 2009, at 10:50 AM, Brian J. Murrell wrote:> On Wed, 2009-04-01 at 10:38 -0500, Robert Olson wrote: >> >> Thank you to all for your suggestions. In particular the one above >> jogs my memory - I had done exactly that in a previous test >> incarnation of a lustre service here by hacking a "modprobe lustre" >> into the ypbind startup script. Having an actual startup script to do >> it is just a little cleaner :-). > > Well, if you are going to take that approach, don''t most (all?) > distros > provide a hook to manually load modules. RHEL5 does. I''d think > simply > adding lnet there would take care of it rather than writing whole > initscripts to do it. > > b. > > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss at lists.lustre.org > http://lists.lustre.org/mailman/listinfo/lustre-discuss
On Wed, 2009-04-01 at 10:58 -0500, Robert Olson wrote:> oh right, of course. lnet is sufficient for triggering what is needed?I believe so. But my point was using what is already there to load modules early rather than writing initscripts to do it. b. -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 197 bytes Desc: This is a digitally signed message part Url : http://lists.lustre.org/pipermail/lustre-discuss/attachments/20090401/65bfb5ee/attachment.bin