|Title: ||Linux: eloqdb6 causes system delay|
|Author: ||Michael Marxmeier (firstname.lastname@example.org)|
Q: When using sync write mode (default with
A.06.20) my Linux server seems to stall every minute.
A: This is caused by an issue of the Linux
kernel syncing large files (with the ext2 file system).
When the eloqdb6 performs in sync write mode (default as
of A.06.20) a fsync() on the data volume(s) is performed
during checkpoint operation (every minute by default).
The algorithm used by the ext2 filesystem for syncing
buffers to disk is inefficient for big files (this is a
know problem but not easy to change). When using the
eloqdb6 in sync write mode (which is the default now)
this may cause delays and the system may seem to stall
for a few seconds when eloqdb6 performs a checkpoint
operation (default every 60 seconds). This gets noticeable
when the database volume files grow beyound 100 MB and
gets worse the bigger the files get.
A kernel patch is available as a workaround.
For more information please refer to the
A.06.20 Release Notes.