Re:How Linux 3.6 Nearly Broke PostgreSQL
by lnanek2 on Thursday October 11, 2012 @11:28AM
(#4641192)
High load can take out entire clusters. Imagine if a change was pushing to NetFlix that slowed it down 20%, the requests would queue up to the point where everything fell over.
[ Parent
| on HN ]
Re:How Linux 3.6 Nearly Broke PostgreSQL
by prodigal_erik on Thursday October 11, 2012 @12:20PM
(#4641486)
That calls for shedding load as triage. If you're beyond capacity, it's better (less bad, anyway) for 20% of requests to fail quickly than to actually attempt them if that will prevent you from completing the other 80%.