Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
/ 2004-06-01 11:29:37 +0200 \ Paolo Morandi: > Lars Ellenberg wrote: > >>- for shutdown the cluster i must before stopping drbd on primary node > >>and after shutdown the two machine (graceful shutdown): a question, but > >>shutting down are not executed the 'drbd stop' ? > > > > > >depending on your setup. most likely it is. > >but you may have a race, if you shutdown both at the same time. > Now the cluster is running with heartbeat and work fine. If i start drbd > i see on the primary machine in Secondary/Secondary state, after > heartbeat i see the filesystem mounted and the cluster state in > Primary/Secondary. > If i stop heartbeat the filesystem are unmounted and the cluster state > back to Secondary/Secondary. So if I stop drbd, can restart without sync. > The problem is when i want shutdown the cluster: if i dont make manually > the operation at the new statup the cluster go to SyncAll. so what do you mean by "shutdown the cluster"? what exactly do you do? what is involved? what might that implicate? answer yourself that questions, and keep in mind that drbd needs to be Con:Sec/Sec when before it loses connection. > >>An other question: > >> > >>are possible to increase the speed of syncronization? For syncing 73Gb > >>needs 20h! So is very difficult to probe the mechanism. > > > > > >73G / 20h == 73*1024*1024 k / 20*3600sec == 1063 k/sec > >this is slooow. > >I have seen sync speeds of >= 50000 k/sec == 50M/sec > > > >so the answer obviously is yes. how you do this depends from your > >particular hardware and configuration. > >double check sync-max please, and whether DMA is enabled for your disks, > >and ... > >just do your homework and find your bottleneck :) > > > I use for each machine two SATA disk on Intel Controller 5013C-T in > softrware RAID 1 > > I have edited the original drbd.conf so, for the disk, i see this records: > sync-min = 500k > sync-max = 1M # maximal average syncer bandwidth > tl-size = 5000 # transfer log size, ensures strict write ordering > timeout = 60 # unit: 0.1 seconds > connect-int = 10 # unit: seconds > ping-int = 10 # unit: seconds > ko-count = 4 # if some block send times out this many times, > > I must increase sync-min to 5M and sync-max to 500M ? there is no must. but use your common sense. you complain about an average sync speed of 1M/sec, but you configure an average sync speed of 1M/sec. so what exactly are you complaining about? :) Lars Ellenberg