(OT) Mailing list server problems

[Bill Campbell]

On Mon, Feb 13, 2017, Fairlight via Filepro-list wrote:
>On Mon, Feb 13, 2017 at 10:20:16AM -0800, Bill Campbell via Filepro-list thus spoke:
>> 
>> The server for our Mailman mailing lists runs in a VMware virtual
>> machine under the old free VMware server on a CentOS Linux box.
>> That machine hadn't come up clean, and I had to fiddle a bit to
>> get things running again.  It wasn't until yesterday afternoon
>> that I figured out that the root file system had been corrupted,
>> and mostly disappeared after I booted off a Knoppix CD and ran a
>> full 'fsck -f -y' on the root file system with everything in the
>> /lost+found directory, largely useless.
>
>What filesystem are you using?  I want to know which to avoid for life.  :)

ext3.

>> At that point, I removed the /lost+found stuff, ran another fsck,
>> then tried to copy the root file system using rsync from the NAS
>> box.  This failed as it said the file system was full, even
>> though the 8GB file system should hold less than 5GB.  I tried
>> again using find and cpio, with the same result.
>
>This is sounding suspiciously like ext4.  You can have 100GB total, and
>only 45GB of it in use, but you can exhaust the inodes if you're doing
>heirarchical storage like Postfix spools, IMAP boxes...anything which
>generates more inodes than space.  Try a `df -i`.
>
>The -really- bad thing is that, due to one of -the- worst 'design'
>decisions in recent history, ext4's inode table is static.  Whatever
>number of inodes created at mkfs, that's what you have for the life of the
>filesystem.  I have a few systems which need to be custom-built by doing
>the install, then syncing everything to a spare filesystem which has enough
>inodes, reworking the filesystem from scratch with a mkfs with a huge inode
>count, then syncing everything back.  All of that is really bitchy under
>LVM, too, depending -whose- recovery/live discs you use.  I actually need
>one CentOS and one SuSE disk.  I could do it all with CentOS, had they
>bothered to include `ls` on the frakkin' rescue disk.  I kid you not, no
>`ls` existed on the one I was using for 6.x when I last hit this issue.
>And SuSE and CentOS have differing ideas on how to address the devices and
>grub, just adding fun to the mix.
>
>I'm starting to suspect you're running ext4, especially as it's the default
>filesystem for CentOS 6.x.

Nope.  I'm running ext4 on a couple of systems, and haven't had
problems so far.

All of our systems use courier-imap which stores in Maildir
format, one file per message.  So far haven't run into issues
with this even with ISPs having over 10,000 accounts.

Bill
-- 
INTERNET:   bill at celestial.com  Bill Campbell; Celestial Software LLC
URL: http://www.celestial.com/  PO Box 820; 6641 E. Mercer Way
Voice:          (206) 236-1676  Mercer Island, WA 98040-0820
Fax:            (206) 232-9186  Skype: jwccsllc (206) 855-5792

The best we can hope for concerning the people at large is that they be
properly armed.  -- Alexander Hamilton, The Federalist Papers at 184-188