Enabling TRIM in Debian fstab for ext4 file-system on Samsung SSD 960 EVO NVMe M.2

So I was trying to find why in my Debian 9 system my SATA drives are called SCSI devices, and I was reading Why my SATA drive is identified as a SCSI device in Device Manager where I read:

The Intel Rapid Storage driver version 12.6 (Released in March 2013) and newer versions classify all drives as SCSI devices for uniformity. This (and later) versions of the driver also allow for TRIM support (Allows for management of data blocks no longer in use) in SSD drives in RAID 0 arrays and other flexibility in operation of storage devices.

So I’d never heard of “TRIM” so I searched for that and found the Wikipedia Trim (computing) article, but Why SSD TRIM Support is So Important and How to Enable It caught my eye, because, “important” you say?

So that article about the importance of TRIM was for Windows, so I searched again and found How to properly activate TRIM for your SSD on Linux: fstrim, lvm and dm-crypt which suggested things were a bit complicated for LVM (and MD RAID?).

I ended up reading How to set up SSD raid and TRIM support? which sent me to Re: Best way (only?) to setup SSD’s for using TRIM which argued that perhaps TRIM configuration wasn’t necessary at all.

It was then I realised that I probably don’t care about TRIM on my MD RAID SSD drives, but I probably do on my M.2 SSD, which isn’t using MD RAID, but which gets massive tgz files written to it and deleted from it every day. So some more searching and I found Samsung SSD 960 EVO NVMe M.2 Review: Ultra Fast, Affordable Storage which said TRIM was supported:

Supporting features: TRIM (Required OS support), Garbage Collection, S.M.A.R.T

So then I found Enable TRIM On SSD (Solid-State Drives) In Ubuntu For Better Performance which showed me how to enable TRIM in /etc/fstab. So the relevant fstab line was:

# /data/fast was on /dev/nvme0n1p1 during configuration
UUID=87bcc5fa-9261-404b-8bc7-a214f4651b49 /data/fast      ext4    noatime,discard 0       2

Note the ‘discard’ option, that’s where the magic happens.

So I unmounted and remounted the partition,

root@tact:/home/jj5# umount /data/fast
root@tact:/home/jj5# mount /data/fast

And dmesg indicated the discard option had been applied:

[34783.251592] EXT4-fs (nvme0n1p1): mounted filesystem with ordered data mode. Opts: discard

Now I guess we wait and see if my performance issues improve…

Fixing mailman can’t discard/defer message via web config

I had a notice from Mailman for one of my lists that there was a message that required administrative attention (i.e. it was spam). The mailman web-interface provides a facility for treating email, however when I went to discard the message it wouldn’t go away. I tried a heap of things and it took a long time to figure out, but eventually I noticed that in the mailman web page the domain name it was using was “intranet.blackbrick.com” whereas that should have been “www.intranet.blackbrick.com”. So I did a little searching about how to configure a list’s base URL and discovered this, which fixed my issue:

/usr/lib/mailman/bin/withlist -l -r fix_url support \
  --urlhost=www.intranet.blackbrick.com

Where ‘support’ above is ‘list name’.