Kent Overstreet - 12.02.24, 21:42:26 CET: [thoughts about whether a cache flush / FUA request with write caches disabled would be a no-op anyway] > > I may test the Transcend XS2000 with BTRFS to see whether it makes a > > difference, however I really like to use it with BCacheFS and I do not > > really like to use LUKS for external devices. According to the kernel > > log I still don't really think those errors at the block layer were > > about anything filesystem specific, but what do I know? > > It's definitely not unheard of for one specific filesystem to be > tickling driver/device bugs and not others. > > I wonder what it would take to dump the outstanding requests on device > timeout. I got some reply back from Transcend support. They brought up two possible issues: 1) Copied to many files at once. I am not going to accept that one. An external 4 TB SSD should handle writing 1,4 TB in about 215000 files, coming from a slower Toshiba Canvio Basics external HD, just fine. About 90000 files was larger files like sound and video files or installation archives. The rest is from a Linux system backup, so smaller files. I likely move those elsewhere before I try again as I do not need these on flash anyway. However if the amount of files or data matters I could never know what amount of data I could write safely in one go. That is not acceptable to me. 2) Power management related to USB port. Cause I am using a laptop. It may have been that the Linux kernel decided to put the USB port the SSD was connected to into some kind of sleep state. However it was a constant rsync based copy workload. Yes, the kernel buffers data and the reads from Toshiba HD should be quite a bit slower than the Transcend SSD could handle the writes. I saw now more than 80-90 MiB/s coming from the hard disk. However I would doubt this lead to pauses of write activity of more than 30 seconds. Still it could be a thing. Regarding further testing I am unsure whether to first test with BTRFS on top of LUKS – I do not like to store clear text data on the SSD – or with BCacheFS plus fixes which are 6.7.5 or 6.8-rc4 in just in the case the flush handling fixes would still have an influence on the issue at hand. First I will have a look on how to see what USB power management options may be in place and how to tell Linux to keep the USB port the SSD is connected to at all times. Let's see how this story unfolds. At least I am in no hurry about it. Best, -- Martin
