Matus UHLAR - fantomas wrote:
True, but this is the point: If you have plenty of swap, there is a much better chance to recognise and fix the problem before a crash.On 22.06 22:47, Hendrik Voigtländer wrote:
With "growing" I mean the squid process increasing over the time. If I read the FAQ correctly this could be caused e.g. if more and more objects go in the the cache_dirs thus increasing the space needed for the index, i.e. if squid starts with a clean cache_dir.
However, my squid is no longer increasing in process size, cache dirs are full & the load is the same all day.
Snapshot from top (idle squid at night). 21603 proxy 9 0 1017M 1.0G 1180 S 0.0 50.3 1:46 squid
It uses roughly 50% of the RAM (machine has 2GB) , the rest is used by other processes and buffers/cache.
I just think it is dangerous to disable swap, if one doesn't know how large the squid process will get, i.e. probably larger than the physical memory and this causing the OS to kill processes randomly (I had this problem with java-stuff eating up all memory).
I do not think so. The same can happen if you have swap or not - it's always used as virtual memory. Having more VM just delays running machina out of it,
which only happens if there's some memory leak in squid,
libraries or other applications running on that machine.
What about a squid with a huge index (large cache_dir/lot of cache objects) and a large cache_mem setting? What happen if squids minimum memory requirements exceed the virtual memory available (swap or not?).
If your squid is not using more memory for longer time, we can assume there's no leak,
Quite sure about that, I use the debian/stable package. I never had a problem with my selfcompiled squid's on solaris as well.
and as long as it only uses half of the RAM, you have
still 1 GB free for processes etc, so you probably do not need swap at all.
That is why I disabled it...
Yes, I remember reading something somewhere about that. Isn't that behaviour tunable somehow?
What puzzles me is that my machine started to use swap at all as plenty of memory is available, that is why I disabled swap with a perfomance boots as a result.
That is feature of VM systems, they store unused data onto swap even if it's not needed. If in any case the mamory use will grow, unused data will not have to be stored on disk because they alredy are, so the system will spare swapping that time.
