Bret Orsburn
2008-Aug-27 19:02 UTC
[Samba] Excessive disk activity from browse.dat regeneration
Greetings, I'm trying to track down and eliminate the sources of excessive disk activity in an idle system that is resulting in premature hard disk failure. Access time updates to inodes turned out to be the worst culprit, triggering writes every 35 seconds or so. Mounting filesystems with the noatime option fixed that problem. But not too far behind inode updates is the frequent regeneration of the browse.dat file by nmbd. My first thought was to move browse.dat to a tmpfs so nmbd could create the file as often as it likes without chewing up our hard disks. But the lock directory that contains browse.dat also contains a bunch of other files and some of them seem to want to be persistent. I started down the path of spinning a web of symlinks to put everything in a place where it will be happy. But there seem to be several different lifecycles represented in this collection of files and making them all happy is looking trickier than I had hoped. This seems like the sort of thing that other people would have figured out by now. I've searched the samba archives and haven't found any discussions on exactly this point. Before I dig deeper into the code, could some of you more experienced Samba hands point me to a work-around for this problem? Thanks. I'm using Samba 3.0.0 on Redhat 7.3. (Yes, I know that's very old.) ---- Bret Orsburn
Possibly Parallel Threads
- wierd ssl-parameters.dat regeneration error
- "Invalid KEX record length" during SPTPS key regeneration and related issues
- DTMF regeneration on PRI
- "Invalid KEX record length" during SPTPS key regeneration and related issues
- [ext] dovecot 2.3.7.2-1~bionic: Performance issues caused by excessive IO to ~/mdbox/mailboxes/INBOX/dbox-Mails/dovecot.index.tmp