2016-04-16 21:44 GMT+02:00 Rob Willett <rob.sqlite at robertwillett.com>:
> If you have a load average of 15 then that normally means you have a
> massively overloaded Linux box. I don?t know your system but I get worried
> around a load average of 3-4 on our boxes. Load Average is a very crude
> measurement but a high number tends to be bad.
>
?Yes, normally it is about 2-3 on my system.
?
> If your CPU is only running at 15% (how do you know this?) then the
> problem is likely elsewhere.
?I am using htop for that.
?
> My first thought is swap space, check what the disk io is, the fact you
> mention Java would immediately make me look at the amount of memory
> allocated to the JVM. For some reason many JVM?s allocate a pitifully small
> amount of memory, Java then runs out of memory pretty quickly and spends
> the next few secs/mins/hours fighting with the OS for resources.
>
?At the moment it is difficult to check, because of the sluggishness, ?
?but luckily I have a service running that puts vmstat info in a SQLite
database. :-D So after the program is finished I can look into the data.?
The Java program I am running does not use swap and also not much memory
(113 MB). I wrote some scripts for getting that kind of information:
https://github.com/CecilWesterhof/BashLibrary/tree/master/bin
Top, netstat, ps etc are your friends here. You need to look at them and
> see whats kicking your load average into the stratosphere. I?d be surprised
> if its Sqlite,
?I think Java is more likely, but it does not hurt to check.
?
> we don?t normally drop tables, but from memory, dropping a 10GB table took
> no time. It might have been seconds or a minute, we didn?t measure it as it
> wasn?t an issue.
?At the moment I am filling the table. After is is filled I will try
dropping it again.
--
Cecil Westerhof
