With cyrus-imap (and mutt) code of 11 November, I just got: IOERROR: mapping cache file for user.prlw1.http: Cannot allocate memory Fatal error: failed to mmap cache file I think that cyrus can work on systems with no mmap, so if mmap fails, could mmap-avoiding methods be used instead (read/write)? (Haven't checked the code...) But there is another side to this which puzzles me. After upgrading the system, I essentially did reconstruct user/prlw1 opened INBOX (about 233,000 messages) using mutt tagged 9651 of them copied the tagged messages to an empty, freshly created user/prlw1/http
During this phase, the imap server failed to mmap the new mailbox's cache file. INBOX with all its messages has -rw------- 1 cyrus mail 347M Nov 13 05:23 cyrus.cache -rw------- 1 cyrus mail 155B Nov 12 10:16 cyrus.header -rw------- 1 cyrus mail 20M Nov 13 05:23 cyrus.index # ls -1 | wc -l 233797 yet new http folder, which should contain far fewer messages has -rw------- 1 cyrus mail 804M Nov 12 21:19 cyrus.cache -rw------- 1 cyrus mail 155B Nov 12 18:56 cyrus.header -rw------- 1 cyrus mail 41M Nov 12 21:19 cyrus.index # ls -1 | wc -l 491568 So what are all these files? # ls -liS 1167478 -rw------- 4 cyrus mail 28342472 Jan 28 2008 113844. 1167478 -rw------- 4 cyrus mail 28342472 Jan 28 2008 260251. 1167478 -rw------- 4 cyrus mail 28342472 Jan 28 2008 383849. 1149545 -rw------- 4 cyrus mail 21925261 Sep 24 2007 245262. 1149545 -rw------- 4 cyrus mail 21925261 Sep 24 2007 368860. 1149545 -rw------- 4 cyrus mail 21925261 Sep 24 2007 98855. 1233627 -rw------- 5 cyrus mail 21795046 Jun 12 16:41 160927. 1233627 -rw------- 5 cyrus mail 21795046 Jun 12 16:41 307334. 1233627 -rw------- 5 cyrus mail 21795046 Jun 12 16:41 430932. 1233627 -rw------- 5 cyrus mail 21795046 Jun 12 16:41 474643. 1172884 -rw------- 4 cyrus mail 20400877 Mar 3 2008 118751. 1172884 -rw------- 4 cyrus mail 20400877 Mar 3 2008 265158. 1172884 -rw------- 4 cyrus mail 20400877 Mar 3 2008 388756. Somehow multiple copies of the same file are being written with different filenames, yet same inode number. Any guesses on what is happening? (I have the telemetry log) (NetBSD-current/i386, ffs+wapbl/raidframe 1 filesystem) Cheers, Patrick
