invalidation by namespace Re: Listing all keys on a server: why?

Don MacAskill don at
Mon Jul 31 02:42:11 UTC 2006

Serhat Sakarya wrote:
> On 7/29/06, Don MacAskill <don at> wrote:
>> If Server A crashes for a few minutes, and Servers B-Z get remapped data,
>> and then Server A recovers and Servers B-Z aren't flushed, then there 
>> exists
>> the likelihood that you'll read bad data from Servers B-Z again if 
>> Server A
>> should fail.
> What do you think about giving the remapped data on servers B-Z a low
> expiration time? This will get them flushed rather quickly. Then stale data
> will only be an issue if server A fails again within that short time 
> period.

Everything always depends on use case, but even a tiny amount of time is 
not acceptable for my use case.  Since it's acting as a buffer for my 
DB, there's a chance the stale, incorrect data could then corrupt my DB 
with invalid data.  If your use case can tolerate stale/corrupt/invalid 
data, then so be it.  :)

As long as you have a decent # of memcached instances for your load, 
there's really very little impact when a single instance dies and you 
don't rebalance.


More information about the memcached mailing list