- Possible alerts
- - bins in a cluster are not matched in size
- - clock deviation within servers (and with the client) outside some limit
- - inconsistent builds between servers
- - variation on connection and answer delay
- - non-zero eviction rate
- - inconsistent slab factors between bins
- - multiple uses of same bin in monitor config
- - rusage_* increase on some servers outside normal variations
- - inconsistent uptimes
- - optimizations: different slabs factor simulation shows less memory use than current
- Suggestions on
- http://code.google.com/p/memcached/wiki/NewServerMaint
|