Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
/ 2004-06-04 08:52:24 +0100
\ Tim Jackson:
> Just to pick up on this: yes, if you would consider mentioning the
> sync-group parameter in the "don't do this" comment about multiple devices
> on the same spindle, it would be less confusing to people new to DRBD
> (which includes me!). As it is, someone reading through the file first
> reads about the sync-group parameter but then, in spite of that, later
> reads "don't do this", which puts them off (well, at least it did for me).
> How about "btw, don't do this unless you have the resources in different
> sync-groups"?
hysterical reasons :)
I simply forgot to alter those other comments when I added the
sync-group parameter.
> So, following on from that, I have a (probably common) situation: two
> nodes, a limited number of physical disk slots and lots of services which
> can run on either node. Each node can handle all the load on its own, but
> as a simple kind of load sharing and to save one machine sitting idle,
> would the below configuration work OK and with no worse performance (as
> far as DRBD syncing speed is concerned) than having 1 big DRBD resource?
> It uses two DRBD resources on the same spindle, but with different
> sync-groups:
yes, sure, as long as you do not forget to put in sync-max (and maybe
sync-min) with some appropriate values, defaults to 250 K/sec!, which is
probably not what you want, and put the sync-group in the net {}
section, where it belongs, and ...
lge