From: W K. <wk...@bn...> - 2011-05-18 19:59:58
|
And resolved. From the DoveCot manual. "By default Dovecot mmap()s the index files. This may not work with all clustered filesystems, and it most certainly won't work with NFS. Setting mmap_disable = yes disables mmap() and Dovecot does its own internal caching. If mmap() is supported by your filesystem, it's still not certain that it gives better performance. Try benchmarking to make sure." which indeed solved the problem. Sorry for the noise. -bill On 5/18/2011 12:51 PM, W Kern wrote: > More info > > This appears to be mmap with DoveCot issue. > > May 18 12:47:49 ariel2 dovecot: IMAP(xxx): mmap() failed with index > file /home/xxx/Maildir/.Archives.2011/dovecot.index: No such device > May 18 12:47:49 ariel2 dovecot: IMAP(xxx): file mail-index.c: line > 1900 (mail_index_move_to_memory): assertion failed: (index->fd == -1) > May 18 12:47:49 ariel2 dovecot: IMAP(xxx): Raw backtrace: imap > [0x80acd51] -> imap [0x80acc6c] -> imap [0x808b317] -> > imap(mail_index_open+0x48e) [0x808cefe] -> > imap(index_storage_mailbox_init+0x152) [0x8083c82] -> imap [0x80654c3] > -> imap [0x80662b2] -> imap(cmd_copy+0x258) [0x8057098] -> > imap(cmd_uid+0x50) [0x805aa00] -> imap [0x805af88] -> imap [0x805b00c] > -> imap(_client_input+0x6c) [0x805b6dc] -> > imap(io_loop_handler_run+0x110) [0x80b29f0] -> imap(io_loop_run+0x1c) > [0x80b1f2c] -> imap(main+0x4c0) [0x8063630] -> > /lib/libc.so.6(__libc_start_main+0xdc) [0x8cbe9c] -> imap [0x8056181] > May 18 12:47:49 ariel2 dovecot: child 5515 (imap) killed with signal 6 > > -bill > > On 5/18/2011 12:19 PM, W Kern wrote: >> >> Greetings. >> >> After extensive testing MFS on a test cluster, we proceeded to deploy >> a smallish (3TB) for a IMAP MailServer (DoveCot). >> >> The system is currently 3 chunkservers with a single 1TB drive, a >> Master and MetaServer, all running CentOS5.x and EXT3 with a default >> install. MooseFS is latest 1.6.20 >> The Goal is 2. >> >> Now, I know MooseFS isn't for small files but we wanted to torture it >> to see how things worked and it turns out performance is very good. >> The users can't tell the difference between the new setup and the old >> setup which was a DRBD active/passive arrangement. In fact, some >> remote users are reporting improved performance. Disk usage is a lot >> more but that was to be expected given the clusters. >> >> We have come across an issue. >> >> If a client uses Thunderbirds 3's ARCHIVE button, the Thunderbird >> program immediately reports losing a connection to the server. This >> does not occur on directed attached storage (with or without DRBD). >> >> The exact message is "Server jo...@fo... <mailto:jo...@fo...> has >> disconnected. The server may have gone down or there may be a network >> problem". In addition, although the 'archived' file has been moved >> to the Archive folder, it is no longer accessible without receiving >> that error and a look at the raw Maildir shows that the folder was >> created ".Archives.2011" >> but there is no message in there. So the archive button will 'lose' >> that email. Attempts to Archive other emails also result in a loss >> of the message. >> >> In their instructions Thunderbird describes the Archive function with >> some of the following details: >> >> * For each account, a folder can be specified to archive messages >> into with a single function. >> * *Note:* Archiving involves a /move/ operation rather than a >> /copy/. >> * The currently implemented scheme creates an Archives folder >> which contains subfolders for each year. In turn, those folders >> /can/ contain per-month subfolders (see figure to the right). >> This is determined by a preference >> *mail.server.default.archive_granularity* (or >> *mail.server.server*#*.archive_granularity* for a specific >> account) with 0=single Archives folder, 1=by-year subfolders >> (default), and 2=year/month subfolders. >> >> >> I can reproduce it on every Thunderbird 3 client irregardless of >> Linux/Mac/Win7 operating system and it occurs every single time. >> >> I will be testing Outlook shortly, but I wanted to throw this out >> there first. >> >> I am looking for ideas as to how to fix this issue or even debug it. >> >> Sincerely, >> >> -bill >> >> >> > |