Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On 05/23/2011 10:28 PM, Richard Stockton wrote: > Yes, it's not really CPU, it's system load and that appears to be caused > by high Wait time, "kernel waiting for I/O completion" (thanks to Pascal > BERTON for that insight). A typical "top" during high load shows this... > > top - 12:59:00 up 12 days, 21:58, 8 users, load average: 11.76, 13.03, > 13.55 > Tasks: 289 total, 1 running, 285 sleeping, 2 stopped, 1 zombie > Cpu(s): 0.2%us, 1.6%sy, 0.1%ni, 20.0%id, 76.6%wa, 0.2%hi, 1.4%si, > 0.0%st > Mem: 4043364k total, 4015656k used, 27708k free, 1822780k buffers > Swap: 8385920k total, 220k used, 8385700k free, 1522456k cached > > So, the question becomes, "what is causing all that I/O wait time? and why > wasn't it happening 2 weeks ago?". Very little has changed in the last 14 > days, other than the load going way up. Interesting. DRBD could conceivably be the cause of this I/O wait. If so, the root cause might be a bad NIC or other network component. Replace if possible. Others on this list can hopefully more helpful in pinpointing such problems. Regards, Felix -- MPeX.net GmbH / Werner-Voß-Damm 62 / D-12101 Berlin / Germany MPeXnetworks / www.mpexnetworks.de Tel: ++49-30-78097 180 / Fax: ++49-30-78097 181 Sitz, Registergericht: Berlin, Amtsgericht Charlottenburg, HRB 76688 Geschäftsführer: Lars Bräuer, Gregor Lawatscheck