tcmu-runner iGW is seen behaving as a generic iSCSI target, seen from vSphere, with dataflow on 1 AO and n ANO's as "hot-standby" for each LUN. Load-balance is therefore limited to be between images, if you use tcmu-runner based deployment. Each LUN will have one AO(Active-Optimized) path and n non-optimized paths. All data for a given LUN will, under normal operations, go through the AO path with the rest being "hot-standby" in case one of the AO assigned GW for the particular LUN goes down. LUN's are afaik. pinned to a given iGW as AO, when the image is created and choosing the iGW to pin an image on, is based on the total number of images that's already pinned to the GW, selecting the one with the least number of pinned images. When you design the iGW deployment, you should also bare in mind that #1 There is a limit on the supported number of iGW's if you use a commercially supported deployment of ceph. #2 There is a performance limit (iops and throughput). If you have ssd based pool's exposed, then the bottleneck will likely be the LUN itself on the iGW, assuming a lacp based setup. For vSphere datastore design considerations, my genreal recommendation is that you make your datastores large'ish(assuming vmfs6 on >=6.5u3), but not so large that you can't pull one out (put datastore in in maintenance mode, in a datastore cluster), this gives flexibility for unforseen issues and vmfs upgrades alike. Also ensure that your SATP, PSP and recovery timeout is set in accordance with the doc's Regarding ceph pools, I generally make one for each device-class that provides rbd. but this depends on your requirements for resiliance etc. on the ceph level. Remember that _general_ recommendations seldom fit any specific deployment well. Regards Heðin Ejdesgaard +298 77 11 10 On Thu, 2019-12-05 at 11:24 -0800, Philip Brown wrote: > Okay then.. how DO you load balance across ceph iscsi gateways? > You said "check the docs", but as far as I can tell, that info isnt > in there. > Or at least not in the logical place, such as the iscsi gateway setup > pages, under > https://docs.ceph.com/docs/master/rbd/iscsi-targets/ > > > ----- Original Message ----- > From: "Paul Emmerich" <paul.emmerich@xxxxxxxx> > To: "Philip Brown" <pbrown@xxxxxxxxxx> > Cc: "ceph-users" <ceph-users@xxxxxxxxxxxxxx> > Sent: Thursday, December 5, 2019 11:08:23 AM > Subject: Re: best pool usage for vmware backing > > No, you obviously don't need multiple pools for load balancing. _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |