Messages in this thread |  | | Date | Mon, 10 Jun 1996 11:15:44 +0000 (GMT) | From | Jeff Johnson <> | Subject | 2.0, loggings, cpu quotas, 2.1 issues, etc. |
| |
I seem to have a problem. My messages file quickly grows to be several megs after a few days of uptime. The average uptime for this system is two to three weeks, and it was just upgraded to Linux 2.0.0:
- bluebox:~$ uname -a - Linux bluebox 2.0.0 #1 Sun Jun 9 19:30:22 GMT 1996 i586 - bluebox:~$ uptime - 10:55am up 14:34, 4 users, load average: 1.94, 1.51, 1.24 - bluebox:~$ date - Mon Jun 10 10:55:57 GMT 1996
This problem existed in 1.3.xx kernels also, but I thought it might go away. It didn't. My /var/adm/messages file contains a few hundred thousand lines of messages such as the following:
- Jun 10 10:31:13 bluebox on Jun 10 10:30:13 GMT 1996 Mon Jun 10 10:31:13 GMT 1996 - Jun 10 10:31:13 bluebox - Jun 10 10:32:13 bluebox on Jun 10 10:31:13 GMT 1996 Mon Jun 10 10:32:13 GMT 1996 - Jun 10 10:32:13 bluebox - Jun 10 10:33:14 bluebox on Jun 10 10:32:13 GMT 1996 Mon Jun 10 10:33:14 GMT 1996 - Jun 10 10:33:14 bluebox - Jun 10 10:34:14 bluebox on Jun 10 10:33:14 GMT 1996 Mon Jun 10 10:34:14 GMT 1996 - Jun 10 10:34:14 bluebox - Jun 10 10:35:14 bluebox on Jun 10 10:34:14 GMT 1996 Mon Jun 10 10:35:14 GMT 1996 - Jun 10 10:35:14 bluebox - Jun 10 10:36:14 bluebox on Jun 10 10:35:14 GMT 1996 Mon Jun 10 10:36:14 GMT 1996 - Jun 10 10:36:14 bluebox - Jun 10 10:37:14 bluebox on Jun 10 10:36:14 GMT 1996 Mon Jun 10 10:37:14 GMT 1996 - Jun 10 10:37:14 bluebox - Jun 10 10:38:14 bluebox on Jun 10 10:37:14 GMT 1996 Mon Jun 10 10:38:14 GMT 1996 - Jun 10 10:38:14 bluebox - Jun 10 10:39:15 bluebox on Jun 10 10:38:14 GMT 1996 Mon Jun 10 10:39:14 GMT 1996 - Jun 10 10:39:15 bluebox - Jun 10 10:40:15 bluebox on Jun 10 10:39:14 GMT 1996 Mon Jun 10 10:40:15 GMT 1996 - Jun 10 10:40:15 bluebox - Jun 10 10:41:15 bluebox on Jun 10 10:40:15 GMT 1996 Mon Jun 10 10:41:15 GMT 1996 - Jun 10 10:41:15 bluebox
And they go on, and on, and on, and on, and on, and on.....
Is this a problem with my configuration?
Also, off of that subject, here are a few things to do/consider now that 2.0 is out:
***** The /*MOST*/ important: A CPU quota system:
This should work a variety of ways. It should have the functionality to allow the sysadmin to limit the total amount of CPU time users can have per login, day, month, etc. Ex: "CPU quota exceeded: 1 hour for login" or similar. I would also like to be able to limit the amount of current CPU time a user can take up by percent. This would have to be a modification to the schedulear code. For example, I could specify that no user can take more than 20% of the CPU time, or that all total users can only take 85% of CPU time. Per user limits would be nice, so that certain users (those who like to fork bomb and run recursive scripts for example) can only have 15% of the CPU time at best, while others who have special needs can use up to 50% for themselves, regardless of other restrictions on their group. I'm sure this could be done, but some people would have to work on it.
***** Second, I would like to see some of the meter-0.2 or Portato code integrated into kernel source and/or modules. This way, system status lights could be updated by the kernel, without reading /proc or calling uptime/ps/the like. This might be considered bloat, but that why it is optional.
***** Profiled/stripped swap. Swap could be written to two drives at once for faster speed. Instead of swapping, say 10MB, to one disk all at once, 5MB could be swapped to 1 SCSI disk, and 5MB to another. You coudl also have profiled swap, where certain swap partitionns/files have priority over others, such as a fast SCSI before swapping to a slower IDE, etc. I believe that work has already been done on this.
***** Better removeable media/cartridge support. Many devices (Zip, Jazz, Bernouli, CDROM, etc) can be ejected by software. When the disk is unmounted, the device should eject it's disk. This should not be very difficult to implement.
***** I use a patch to allieviate some interrupts that are not needed on standard equipment. It might be good to put this in as a kernel configure option for 2.1. The patch is included with this file.
***** SMP support should be a standard kernel configure option in 2.1.
***** Lastly, the Rob Krawitz P5/FPU memcpy() should be included into 2.1. It has been working flawlessly since 1.3.8x, and I have never had any bit flips or corruption of any kind. I can do a make -j on the kernel souces without a problem. His machine is rather flaky, and the only other problem is that SMP support together with pentium memcpy will not boot.
-- Jeff Johnson GCS d- s: !a C+++ UA++(+++) P+ L+ trn@gate.net E---- W+++ N+++(+++++) K- w(+) O(-) KE4QWX M- V-(--) PS+ PE Y++ PGP+++(+++++) t- http://www.gate.net/~trn 5 X+++(+++++) R tv+ b++ DI-- D G++ e* !h r y? Nerdity Test = 66% Hacker Test = 45% 1024/3397E001 1995/06/10 5B 92 8B 34 84 E9 42 26 DC FB F7 C4 1E 0E 80 29
|  |