On Sat, Sep 7, 2019 at 7:47 AM Paul Emmerich <paul.emmerich@xxxxxxxx> wrote: > > On Sat, Sep 7, 2019 at 12:55 AM William Ferrell <willfe@xxxxxxxxx> wrote: > > > > On Fri, Sep 6, 2019 at 6:37 PM Peter Woodman <peter@xxxxxxxxxxxx> wrote: > > > > > > 2GB ram is gonna be really tight, probably. > > > > Bummer. So it won't be enough for any Ceph component to run reliably? > > 2 GB is tough for an OSD. The main problem is recovery. If you attempt > this: the trick is the use fewer PGs per OSD than recommended. Interesting. I take it that reduces the memory requirements? What are the consequences of using fewer PGs than recommended? On a related note, I came across this hardware while searching around on this topic: https://ambedded.com/ambedded_com/ARM_Server_platform.html They position these systems as easy-to-deploy Ceph systems. I'm afraid to even ask the manufacturer how much those things cost, but I noticed the specs of the individual nodes aren't really much different from the HC2, especially the "Mars 200" model. In that model, each node has a dual core CPU and only 2GB of RAM. The 400 model changes the CPU type (but stays dual-core) and bumps the memory up to 4GB. So would the individual nodes in those machines have the same limitations? I guess the chassis provides an internal network fabric so the nodes can talk to each other super fast, so maybe that makes a difference. BTW thanks everyone who's responded so far! You've all been really helpful clearing up my newbie questions :) _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |