Hello I got this error message yesterday evening for a user I've checked LDA error log it happened 6 times in 3 days for 5 differents users, not much but I wonder what's going on as actually the server is moderately loaded ( Easter holidays ) we have currently 250/300 IMAP processes , in "normal" production time we have approx 1000 IMAP processes running. Apr 14 19:24:37 imap(xxxxxxx): Error: FETCH [] for mailbox Drafts (Hugues at ESIEE) UID 17836 got too little data: 16384 vs 18012 Apr 14 19:24:37 imap(xxxxxxx): Error: Corrupted index cache file /var/cache/dovecot/index/xxxxxxx/.imap/Drafts (Hugues at ESIEE)/dovecot.index.cache: Broken virtual size for mail UID 17836 The machine is supposed to be HUDGE for our 4000 mailboxes HP Proliant DL380 G7 , 12 Gb RAM 24 cores !!! I've located index/cache files on another disk than users mailboxes mailboxes are living on a 2Tb RAID array ( 50% full ) index/cache are living on another single disk ( 10% full ) I run Dovecot 2.0.11 + dovecot2-pigeonhole compiled from FreeBSD 8.2 ports Dovecot or hardware problem ??? Thanks a lot for any ideas/infos
On 2011-04-15 2:32 AM, Frank Bonnet wrote:> Dovecot or hardware problem ??? > > Thanks a lot for any ideas/infosSorry, crystal ball is broken, rather than making us guess, maybe you could provide your config details (dovecot -n output for starters)? -- Best regards, Charles
On Fri, 2011-04-15 at 08:32 +0200, Frank Bonnet wrote:> Apr 14 19:24:37 imap(xxxxxxx): Error: FETCH [] for mailbox Drafts > (Hugues at ESIEE) UID 17836 got too little data: 16384 vs 18012 > Apr 14 19:24:37 imap(xxxxxxx): Error: Corrupted index cache file > /var/cache/dovecot/index/xxxxxxx/.imap/Drafts (Hugues at > ESIEE)/dovecot.index.cache: Broken virtual size for mail UID 17836mboxes are a bit annoying things. I guess there's a bug somewhere in Dovecot code, but unless you can show me how to reproduce it I don't think I can do anything about it. (I've earlier spent days looking at the mbox code without seeing any bugs in it.) Maybe once everything else is working perfectly I'll just try once more to rewrite the mbox code to be simpler to understand and become perfect. :)