Re: [lustre-discuss] Lustre and OFED
Hi Eli, we are running lustre without OFED on debian client and server. With lustre 2.4.0 on client and servers no problem at all since years. With lustre 2.5.3 on servers and 2.6. 92 no problems at least for monthes. with lustre 2.5.3 on servers and 2.7 on clients allways ib connection loss. Here I'm wondering if a more recent OFED version could help? We are mostly interested in a rock solid lustre version, lustre 2.6 is fast enough for us, but has a memory leak caused by cache usage, lustre 2.7 was perfect for us in tests with a small number of machines, but fails completly for the full cluster and/or certain tasks. Best Harald On Donnerstag, 27. Juli 2017 22:55:33 CEST E.S. Rosenberg wrote: > Hi all, > > How 'needed' is OFED for Lustre? In the LUG talks it is mentioned every > once in a while and that got me thinking a bit. > > What things are gained by installing OFED? Performance? Accurate traffic > reports? > > Currently I am using a lustre system without OFED but our IB hardware is > from the FDR generation so not bleeding edge and probably doesn't need OFED > because of that > > Thanks, > Eli > > Tech specs: > Servers: CentOS 6.8 + Lustre 2.8 (kernel from Lustre RPMs) > Clients: Debian + kernel 4.2 + Lustre 2.8 > IB: ConnectX-3 FDR ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
Re: [lustre-discuss] Lustre and OFED
Jeff (and Grigory - offlist), Thanks for your fast replies! On Fri, Jul 28, 2017 at 12:09 AM, Jeff Johnson < jeff.john...@aeoncomputing.com> wrote: > Eli, > > The biggest driver is usually the drivers. Newer Mellanox hardware not yet > supported, or supported well, by kernel IB. Way back in the days of old > there were some interoperability issues where everything (clients and > servers) needed to be the same drivers and libraries but much of that was > cleaned up. There could be situations where OFED is needed on the server > side to support something under the Lustre layer like OST or MDT block > devices via iSER, SRP, NVMeF, etc. > > There may be other reasons but those are off the top of my head. > So currently everything seems to be working just fine without OFED, my only complaint is that the normal Linux interface counters don't report traffic properly which means I have to write my own perfquery wrappers for tools like zabbix etc. I may try adding OFED if I have time at some point but I hope by then to at least have moved our servers to CentOS 7.3 + Lustre 2.9/10. Has anyone ever run benchmarks of vanilla vs. OFED? Thanks again, Eli > > --Jeff > > On Thu, Jul 27, 2017 at 4:55 PM, E.S. Rosenberg < > esr+lus...@mail.hebrew.edu> wrote: > >> Hi all, >> >> How 'needed' is OFED for Lustre? In the LUG talks it is mentioned every >> once in a while and that got me thinking a bit. >> >> What things are gained by installing OFED? Performance? Accurate traffic >> reports? >> >> Currently I am using a lustre system without OFED but our IB hardware is >> from the FDR generation so not bleeding edge and probably doesn't need OFED >> because of that >> >> Thanks, >> Eli >> >> Tech specs: >> Servers: CentOS 6.8 + Lustre 2.8 (kernel from Lustre RPMs) >> Clients: Debian + kernel 4.2 + Lustre 2.8 >> IB: ConnectX-3 FDR >> >> ___ >> lustre-discuss mailing list >> lustre-discuss@lists.lustre.org >> http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org >> >> > > > -- > -- > Jeff Johnson > Co-Founder > Aeon Computing > > jeff.john...@aeoncomputing.com > www.aeoncomputing.com > t: 858-412-3810 x1001 <(858)%20412-3810> f: 858-412-3845 > <(858)%20412-3845> > m: 619-204-9061 <(619)%20204-9061> > > 4170 Morena Boulevard, Suite D - San Diego, CA 92117 > > High-Performance Computing / Lustre Filesystems / Scale-out Storage > ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
Re: [lustre-discuss] Lustre and OFED
Eli, The biggest driver is usually the drivers. Newer Mellanox hardware not yet supported, or supported well, by kernel IB. Way back in the days of old there were some interoperability issues where everything (clients and servers) needed to be the same drivers and libraries but much of that was cleaned up. There could be situations where OFED is needed on the server side to support something under the Lustre layer like OST or MDT block devices via iSER, SRP, NVMeF, etc. There may be other reasons but those are off the top of my head. --Jeff On Thu, Jul 27, 2017 at 4:55 PM, E.S. Rosenbergwrote: > Hi all, > > How 'needed' is OFED for Lustre? In the LUG talks it is mentioned every > once in a while and that got me thinking a bit. > > What things are gained by installing OFED? Performance? Accurate traffic > reports? > > Currently I am using a lustre system without OFED but our IB hardware is > from the FDR generation so not bleeding edge and probably doesn't need OFED > because of that > > Thanks, > Eli > > Tech specs: > Servers: CentOS 6.8 + Lustre 2.8 (kernel from Lustre RPMs) > Clients: Debian + kernel 4.2 + Lustre 2.8 > IB: ConnectX-3 FDR > > ___ > lustre-discuss mailing list > lustre-discuss@lists.lustre.org > http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org > > -- -- Jeff Johnson Co-Founder Aeon Computing jeff.john...@aeoncomputing.com www.aeoncomputing.com t: 858-412-3810 x1001 f: 858-412-3845 m: 619-204-9061 4170 Morena Boulevard, Suite D - San Diego, CA 92117 High-Performance Computing / Lustre Filesystems / Scale-out Storage ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
[lustre-discuss] Lustre and OFED
Hi all, How 'needed' is OFED for Lustre? In the LUG talks it is mentioned every once in a while and that got me thinking a bit. What things are gained by installing OFED? Performance? Accurate traffic reports? Currently I am using a lustre system without OFED but our IB hardware is from the FDR generation so not bleeding edge and probably doesn't need OFED because of that Thanks, Eli Tech specs: Servers: CentOS 6.8 + Lustre 2.8 (kernel from Lustre RPMs) Clients: Debian + kernel 4.2 + Lustre 2.8 IB: ConnectX-3 FDR ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
Re: [lustre-discuss] High MDS load, but no activity
Hi Kevin, On Thu, Jul 27, 2017 at 08:18:04AM -0400, Kevin M. Hildebrand wrote: >We recently updated to Lustre 2.8 on our cluster, and have started seeing >some unusal load issues. >Last night our MDS load climbed to well over 100, and client performance >dropped to almost zero. >Initially this appeared to be related to a number of jobs that were doing >large numbers of opens/closes, but even after killing those jobs, the MDS >load did not recover. > >Looking at stats in /proc/fs/lustre/mdt/scratch-MDT/exports showed >little to no activity on the MDS. Looking at iostat showed almost no disk >activity to the MDT (or to any device, for that matter), and minimal IO wait. >Memory usage (the machine has 128GB) showed over half of that memory free. sounds like VM spinning to me. check /proc/zoneinfo, /proc/vmstat etc. do you have zone_reclaim_mode=0? that's an olde, but important to have set to zero. sysctl vm.zone_reclaim_mode failing that (and assuming you have a 2 or more numa zone server) I would guess it's all the zone affinity stuff in lustre these days. you can turn most of it off with a modprobe option options libcfs cpu_npartitions=1 what happens by default is that a bunch of lustre threads are bound to numa zones and preferentially and agressively allocate kernel ram in those zones. in practice this usually means that the zone where IB card is physically attached fills up, and then the machine is (essentially) out of ram and spinning hard trying to reclaim, even though all the ram in the other zone(s) is almost all unused. I tried to talk folks out of having affinity on by default in https://jira.hpdd.intel.com/browse/LU-5050 but didn't succeed. even if it wasn't unstable to have affinity on, IMHO having 2x the ram available for caching on the MDS and OSS's is #1, and tiny performance increases from having that ram next to the IB card is a distant #2. cheers, robin >I eventually ended up unmounting the MDT and failing it over to a backup >MDS, which promptly recovered and now has a load of near zero. > >Has anyone seen this before? Any suggestions for what I should look at if >this happens again? > >Thanks! >Kevin > >-- >Kevin Hildebrand >University of Maryland, College Park >Division of IT >___ >lustre-discuss mailing list >lustre-discuss@lists.lustre.org >http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
[lustre-discuss] LAD'17 - Last call!
Hello Lustre community! These are the last days to send abstracts for LAD'17, do not wait! We will be pleased to hear about your Lustre experiences. Not only developers but also from sites or admins presenting their Lustre deployment and experiences with it. https://easychair.org/conferences/?conf=lad17 *ABOUT* EOFS and OpenSFS are happy to announce the 7th LAD will be held in Paris, France, at Salon des Arts et Métiers! This will be a 2-day event, from 4th to 5th of October, 2017. This will be a great opportunity for worldwide Lustre administrators and developers to gather and exchange their experiences, developments, tools, good practices and more. **https://www.eofs.eu/events/lad17 *CALL FOR PAPERS * We invite community members to send proposals for presentation. Those talks could cover, by example, tools, administration experiences, configuration setup, developments, tunings, tweaking, ... No proceeding is required, just an abstract for a 30-min technical presentation (including Q). Please send your abstracts using EasyChair, before July 31th, 2017. https://easychair.org/conferences/?conf=lad17 *REGISTRATION *Registration for the workshop is open: http://lad.eofs.eu/ * DATE CHANGE!* Important! Please note that LAD dates have shifted by one day from what was originally announced! LAD will now take place from October 4th to October 5th, which means from Wednesday to Thursday. *SOCIAL EVENT* On Wednesday evening, a dinner will take place at the top of Tour Montparnasse, Paris highest skyscraper! A limited number of spouses can also attend (on a first-come, first-served basis). *SPONSORS* We are very pleased this event is organized thanks to the following generous sponsors: ATOS, CEA, DDN, INTEL and SEAGATE For any other information, please contact l...@eofs.eu ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
[lustre-discuss] High MDS load, but no activity
We recently updated to Lustre 2.8 on our cluster, and have started seeing some unusal load issues. Last night our MDS load climbed to well over 100, and client performance dropped to almost zero. Initially this appeared to be related to a number of jobs that were doing large numbers of opens/closes, but even after killing those jobs, the MDS load did not recover. Looking at stats in /proc/fs/lustre/mdt/scratch-MDT/exports showed little to no activity on the MDS. Looking at iostat showed almost no disk activity to the MDT (or to any device, for that matter), and minimal IO wait. Memory usage (the machine has 128GB) showed over half of that memory free. I eventually ended up unmounting the MDT and failing it over to a backup MDS, which promptly recovered and now has a load of near zero. Has anyone seen this before? Any suggestions for what I should look at if this happens again? Thanks! Kevin -- Kevin Hildebrand University of Maryland, College Park Division of IT ___ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org