This behavior is symptomatic of memory fragmentation.  When reads are
faster than writes, system buffers get filled up, memory fragmentation
gets worse, and things can drag to a crawl.  Check fragmentation by cat
/proc/buddyinfo and see the situation.  Manual attempts to change the
dirty bytes/ratio for a fix may help certain situations, but slowing
down the data passing to the buffers by requesting a hugh amount may be
enough to force the defrag to provide the user requested buffersize. for
tar use the --record-size=500M to force the defrag to trigger.  Of
course, it also helps to reduce buffer use with nocache on the copy
command and to turn off daemons caching things like .img files (virt-
manager has such a daemon). The copy rate I saw copying a 300GB
partition to a USB3 mounted SSD with just nocache and record-size=500M
on a tar command was 200MB/sec (16 GB memory). Same machine without
mitigations has show rates of 10MB/sec on copying a 60GB file.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2095292

Title:
  rsync to usb ssd blocked for more than 122 seconds

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux-signed-hwe-6.8/+bug/2095292/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to