Hello:
Our scenario is the following:
One web server runninng kernel 2.4.9-e-12(smp), it is used to host an Apache webserver with an integrated application server. The system has 2Gb of RAM and 2Gb of swap. The last day an strange thing happened. We tried to connect to our system using ssh and we can´t, after many attempts we could connect to the system and saw that we couldn´t connect because there was a lack of memory. Every proccess you tried to start failed and gave the following error message" fork failed: Can´t allocate memory (errno=12)".
The output of the free -k command showed that the swap space(there was only 5 Mb of free RAM) wasn´t being used and we don´t understand why. We have found a lot of similar cases of systems that couldn´t fork but were plenty of unused swap space so we thought we would find answers for this problem but we haven´t found any(although we have looked for one a lot).
We think that perhaps our freepages settings are too low( 1.6-4.5-7.4 Mb) and if we merge this low values with an excesive fragmented memory this could explain our memory squeeze problems. Some of us have proposed the theory that everytime a proccess is started it needs a little quantity of contiguous memory and that if the system can´t provide it to the process it dies before the fork is completed. Any idea about what can be happening here?
We have seen that kernel 2.4.9.e-24 seems to fix this kind of problems, what do you think?
Apart from kernel upgrading we want to change the values(freepages) that control the swapping but we have seen in redhat release-notes that these parameters shouldn´t been changed(although it seems more a microsoft disclaimer"You are a fool so please don´t do anything with your system" than a real warning because there aren´t reasons about how can it spoil your system).
Regards,
Mario.
Our scenario is the following:
One web server runninng kernel 2.4.9-e-12(smp), it is used to host an Apache webserver with an integrated application server. The system has 2Gb of RAM and 2Gb of swap. The last day an strange thing happened. We tried to connect to our system using ssh and we can´t, after many attempts we could connect to the system and saw that we couldn´t connect because there was a lack of memory. Every proccess you tried to start failed and gave the following error message" fork failed: Can´t allocate memory (errno=12)".
The output of the free -k command showed that the swap space(there was only 5 Mb of free RAM) wasn´t being used and we don´t understand why. We have found a lot of similar cases of systems that couldn´t fork but were plenty of unused swap space so we thought we would find answers for this problem but we haven´t found any(although we have looked for one a lot).
We think that perhaps our freepages settings are too low( 1.6-4.5-7.4 Mb) and if we merge this low values with an excesive fragmented memory this could explain our memory squeeze problems. Some of us have proposed the theory that everytime a proccess is started it needs a little quantity of contiguous memory and that if the system can´t provide it to the process it dies before the fork is completed. Any idea about what can be happening here?
We have seen that kernel 2.4.9.e-24 seems to fix this kind of problems, what do you think?
Apart from kernel upgrading we want to change the values(freepages) that control the swapping but we have seen in redhat release-notes that these parameters shouldn´t been changed(although it seems more a microsoft disclaimer"You are a fool so please don´t do anything with your system" than a real warning because there aren´t reasons about how can it spoil your system).
Regards,
Mario.