Bug#606016: indexing takes a minute using all CPU and creates 100mb data
I have the same problem, and it makes me wonder.. Who has 100MB
addresses and PIM-data anyway? Perhaps some default setting is a
little bit larger than strictly neccessary (and efficient)?
I don't know what triggers akonadi at login. Perhaps it is related
to nepomukservices also consuming 100% CPU for a long time?
An equaly large problem for me is that those 100MiB are database log
files are invalidated when the user log out and akonadi stops. New
files are created at each login, wasting huge amounts of space in my
nilfs2 based backup. Those logs should be deleted when akonadi close
the database, and hopefully get missed by my nightly backup.
I have mitigated the problem by reducing innodb_log_file_size to 4M,
and trimmed a few other settings as well. 4 MB is probably still
overkill for most users. I don't know what a sensible setting would
be, but suspect many defaults can be trimmed a lot to achieve better
efficiency for most users.