Re: [HPDD-discuss] Clean shutdown

Thanks for the tips Everything is ok, I just have, and I don't know why, a connection listed as 0@lo when running lshowmount Le 07/07/2015 09:27, Martin Hecht a écrit : > maybe this is what you are looking for: > > during recovery on the MDS: > cat /proc/fs/lustre/mds/lustrefs-MDT0000/recovery_status" > shows the number of clients already connected and the number the > server expects. /proc/fs/lustre/mdt/lustre-MDT0000/recovery_status > > during normal operation > cat /proc/fs/lustre/mds/lustrefs-MDT0000/num_exports > shows the number of exports (If I recall correctly, the MGS is a > client as well in this context) > /proc/fs/lustre/mgs/MGS/num_exports /proc/fs/lustre/mdt/lustre-MDT0000/num_exports > > On 07/06/2015 05:16 PM, Scott Nolin wrote: >> the 'lshowmount' command on the servers can do this. >> >> However I sometimes have seen where lshowmount doesn't show any >> connected clients, but they obviously still are. So maybe there's a >> lower-level way than that utility that's better. >> >> Scott >> >> On 7/6/2015 10:09 AM, Jérôme BECOT wrote: >>> That's how i proceed, but i had kernel crashes on the MDS so I >>> unmounted >>> and remounted it and it recovers >>> >>> No way to check wether Lustre servers considers any client still >>> connect >>> or not ? >>> >>> Le 04/07/2015 04:43, Kurt Strosahl a écrit : >>>> Hi, >>>> >>>> I just did a clean shutdown/startup of two lustre file systems >>>> last weekend (also for electrical work). The order that I've found >>>> most effective is to first shutdown all the clients, then unmount the >>>> osts/shutdown the oss systems, then unmount the mdt/mgs and shut down >>>> that system. On the up I've found that turning on the mdt/mgs >>>> system, >>>> letting it mount up. Then mounting the osts (my oss systems don't >>>> mount lustre on boot). Once the last ost to be mounted looks like it >>>> has finished any recovery it needs to do I mount a single client and >>>> make sure it can see the entire system... and then go from there with >>>> the rest of the clients. >>>> >>>> Since all the lustre clients I have are in the same room as the >>>> lustre >>>> file system itself I don't have to check for stray mounts (I just >>>> walk >>>> around the room looking for systems that are still on, and then power >>>> them off). >>>> >>>> Good luck, >>>> Kurt J. Strosahl >>>> System Administrator >>>> Scientific Computing Group, Thomas Jefferson National Accelerator >>>> Facility >>>> >>>> >>>> Message: 1 >>>> Date: Fri, 03 Jul 2015 13:49:32 +0200 >>>> From: J?r?me BECOT <jerome.becot(a)inserm.fr&gt; >>>> To: HPDD-discuss(a)ml01.01.org >>>> Subject: [HPDD-discuss] Clean shutdown >>>> Message-ID: <559676CC.1060009(a)inserm.fr&gt; >>>> Content-Type: text/plain; charset=utf-8; format=flowed >>>> >>>> Hi there, >>>> >>>> We are running a small lustre system (1MGS/MDS + 2 OSTs). I haven't >>>> found the information, so i ask you : >>>> >>>> What is the good order to shutdown the whole system (for electrical >>>> maintenance ...) when you run a combined MGT/MDT : >>>> - disconnect the clients >>>> - is there any way to check if there is a client still connected ? >>>> - unmount the MGT/MDT >>>> - unmount the OSTs >>>> - ... poweroff all the servers >>>> - mount the OSTs >>>> - then the MGT/MDT >>>> - then the clients ? >>>> >>>> or should I stop the ost first and mount them last ? >>>> >>>> Thanks >>>> >>>> >>> >> >> >> >> >> _______________________________________________ >> HPDD-discuss mailing list >> HPDD-discuss(a)lists.01.org >> https://lists.01.org/mailman/listinfo/hpdd-discuss > > > > > _______________________________________________ > HPDD-discuss mailing list > HPDD-discuss(a)lists.01.org > https://lists.01.org/mailman/listinfo/hpdd-discuss _______________________________________________ HPDD-discuss mailing list HPDD-discuss(a)lists.01.org https://lists.01.org/mailman/listinfo/hpdd-discuss