At 21 Oct 2008 15:13:21 +0100, Chris Thompson <[EMAIL PROTECTED]> wrote:
> >I guess you mean 9.5.0-P2, and if so, do you explicitly specify > >max-cache-size? If not, that may be an implicit reason for the > >trouble. The default max-cache-size of 9.5.0-P2 is 32MB, way too > >small for reasonably busy servers. Depending on your available > >memory, I'd suggest specifying at least 256MB for a busy server. > > Would having too small a cache really cause sporadic SERVFAILs? Could be, though the mechanism is indirectly and wouldn't be obvious. > I thought that 9.5.x's super new cache cleaning algorithms were > meant to prevent the long dead periods of previous versions. That's correct, but a super new feature also tends to have a regression bug:-) This is a kind of such a bad side effect (which we can and will fix). > >Aside from that, we're currently fixing a bug in cache entry > >management of 9.5. This may also be related to your symptom, and if > >so, it will hopefully be fixed in the next beta (or RC) version of > >9.5.1. > > Fix number? (so that we will recognise it in the CHANGES file in > due course!) The fix is currently under internal review, and hasn't been assigned a fix number yet. It will however be identifiable by the corresponding bug report number, which is 17628. BTW, if you or other people are particularly interested in the patch, I'll temporarily put it somewhere. It would be good to know the patch will fix actual problems. --- JINMEI, Tatuya
