On 2011-10-31 10:26, James Smith wrote: > I think I have discovered the problem. > > The secondary server, has a degraded array. So it's not in an optimal state.
That the array is degraded should have no impact on performance. If it's degraded _and rebuilding_, DRBD and the resync process compete for I/O, and then a performance hit is inevitable. However, it may be that a much deeper issue is causing both your performance problems _and_ the RAID degradation. But that's something that's near-impossible to ascertain without logging into the box. > This certainly explains the odd IO drop outs I was encountering, where disk > busy would be 100% but with no read or write activity for 20 seconds. > > But, it is a bit odd slow IO on the secondary would cause the master, which > has no degraded array, to also slow down, No, that's expected. http://fghaas.wordpress.com/2007/06/22/performance-tuning-drbd-setups/ That post is a bit dated, but the first few paragraphs still apply to your issue fairly well. > especially when running Protocol A in drbd? The asynchronicity that Protocol A offers is essentially limited to the size of your network send buffer. If your secondary is so slow that it just can't write what comes in over the wire, those buffers are going to fill up and write I/O will slow down. Cheers, Florian -- Need help with DRBD? http://www.hastexo.com/now _______________________________________________ Linux-HA mailing list Linux-HA@lists.linux-ha.org http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems