Howdy y'all My system: Athlon XP 1700+, 512M RAM, 30G HDD, DVD, CD-RW (partly for information partly to brag...) Well, I finally got around to starting to diagnose a problem that's been bugging me since I got this PC. It only has SuSE 8.0 Pro on it, and it's on 24/7 (of course, I mean who ever reboots?) The thing is I've noticed that when I come to it in the morning, I've normally lost about 200M of available RAM (ie free shows about 200M less as free on the -/+ buffers/cache line) This RAM does not become available again at any point, and top doesn't show anything consuming a load of RAM... The same thing happens of course the following night, and you can probably work out where this is going... So, yesterday I added a cron job to run free -m every hour on the hour overnight. The midnight output was: total used free shared buffers cached Mem: 501 407 93 0 64 288 -/+ buffers/cache: 55 446 Swap: 1027 0 1027 And the 1am was: total used free shared buffers cached Mem: 501 477 23 0 186 20 -/+ buffers/cache: 270 230 Swap: 1027 0 1027 At 2pm, the buffers/cache free entry was 232, where it remained for the rest of the night, so I figure it must be something in cron.daily that's causing this, the contents of my cron.daily is: weasel:~ # ls /etc/cron.daily/ . logrotate suse.de-clean-tmp .. medusa.cron suse.de-clean-vi clean_catman suse.de-backup-rc.config suse.de-cron-local clean_core suse.de-backup-rpmdb tetex do_mandb suse.de-check-battery updatedb Does anyone know of any of these that may cause this behaviour? Maybe someone else has had the same thing happen to them and sorted it? Just thought I'd ask before I spend the best part of my weekend trying to resolve this... Cheers James -- James Ogley, Unix Systems Administrator, Pinnacle Insurance Plc james.ogley@pinnacle.co.uk www.pinnacle.co.uk +44 (0) 20 8731 3619 Using Free Software since 1994, running GNU/Linux (SuSE 8.0) Updated GNOME RPMs for SuSE Linux: www.usr-local-bin.org ********************************************************************** CONFIDENTIALITY.This e-mail and any attachments are confidential and may also be privileged. If you are not the named recipient, please notify the sender immediately and do not disclose the contents to another person, use it for any purpose, or store or copy the information in any medium. Any views expressed in this message are those of the individual sender, except where the sender specifically states them to be the views of Pinnacle Insurance plc. If you have received this email in error please immediately notify the Pinnacle Helpdesk on +44 (0) 20 8207 9555. This footnote also confirms that this email message has been swept by MIMEsweeper for the presence of computer viruses. www.mimesweeper.com **********************************************************************
James Ogley said:
Does anyone know of any of these that may cause this behaviour? Maybe someone else has had the same thing happen to them and sorted it? Just thought I'd ask before I spend the best part of my weekend trying to resolve this...
Check the thread from a few days ago about updatedb. It looks like a real problem somewhere in that. //Anders
Check the thread from a few days ago about updatedb. It looks like a real problem somewhere in that.
Cheers, had a look over it, so it seems people agree that there's a memory leak problem, because it's not just the 'free' memory, but the available memory that's reduced, which is what I was saying in my mail. Peter Kleiweg mentioned a program he'd written which with funky use of malloc() frees up memory, a neat hack to fix it (Peter any chance you can pop the code somewhere?) However, seems to me that there must be a better solution, anyone know if there's a patch for this (GNU findutils hasn't been updated since November 1994...) -- James Ogley, Unix Systems Administrator, Pinnacle Insurance Plc james.ogley@pinnacle.co.uk www.pinnacle.co.uk +44 (0) 20 8731 3619 Using Free Software since 1994, running GNU/Linux (SuSE 8.0) Updated GNOME RPMs for SuSE Linux: www.usr-local-bin.org ********************************************************************** CONFIDENTIALITY.This e-mail and any attachments are confidential and may also be privileged. If you are not the named recipient, please notify the sender immediately and do not disclose the contents to another person, use it for any purpose, or store or copy the information in any medium. Any views expressed in this message are those of the individual sender, except where the sender specifically states them to be the views of Pinnacle Insurance plc. If you have received this email in error please immediately notify the Pinnacle Helpdesk on +44 (0) 20 8207 9555. This footnote also confirms that this email message has been swept by MIMEsweeper for the presence of computer viruses. www.mimesweeper.com **********************************************************************
James Ogley said:
Check the thread from a few days ago about updatedb. It looks like a real problem somewhere in that.
Cheers, had a look over it, so it seems people agree that there's a memory leak problem, because it's not just the 'free' memory, but the available memory that's reduced, which is what I was saying in my mail.
Peter Kleiweg mentioned a program he'd written which with funky use of malloc() frees up memory, a neat hack to fix it (Peter any chance you can pop the code somewhere?)
However, seems to me that there must be a better solution, anyone know if there's a patch for this (GNU findutils hasn't been updated since November 1994...)
If there's a problem I think it's in the kernel's VM code. After all, that's what's responsible for freeing up code that processes don't free up themselves. On a side note, if Peter's code works, then there is no memory problem as such. At least not if I understood what he did correctly. If just malloc()ing memory clears it up, then the problem is in how "free" reports memory usage. I don't have systems running long enough, so I don't know if this deteriorates badly enough that it starts swapping and requires a reboot. Does it? //Anders
If there's a problem I think it's in the kernel's VM code. After all, that's what's responsible for freeing up code that processes don't free up themselves.
Good point, and of course VM systems is a serious religious war these days - which one does the Marcelo tree use again? ;-)
On a side note, if Peter's code works, then there is no memory problem as such. At least not if I understood what he did correctly. If just malloc()ing memory clears it up, then the problem is in how "free" reports memory usage. I don't have systems running long enough, so I don't know if this deteriorates badly enough that it starts swapping and requires a reboot. Does it?
I have to reboot the machine in question every couple of days (I actually now have a cron job to do it each morning shortly after 5am) cos otherwise after a couple of days it does start swapping, and it slows down quite horribly. My other two machines at home don't experience this at all - one is my firewall, and so doesn't have findutils-locate installed. The other is running 7.2, so although it does have updatedb, it's a different kernel (2.4.4 as opposed to 2.4.18 on 8.0) so it seems to be something that changed between those kernel versions, was it then that the VM system was changed? Or was it changed in the -aa tree in that time (cos Hubert's kernels are based on the -aa tree IIRC) -- James Ogley, Unix Systems Administrator, Pinnacle Insurance Plc james.ogley@pinnacle.co.uk www.pinnacle.co.uk +44 (0) 20 8731 3619 Using Free Software since 1994, running GNU/Linux (SuSE 8.0) This email was created and sent with Ximian Evolution 1.0.8 NEW: Advogato diary at www.advogato.org/person/riggwelter *********************************************************************** CONFIDENTIALITY. This e-mail and any attachments are confidential and may also be privileged. If you are not the named recipient, please notify the sender immediately and do not disclose the contents to another person, use it for any purpose, or store or copy the information in any medium. Any views expressed in this message are those of the individual sender, except where the sender specifically states them to be the views of Pinnacle Insurance Plc. If you have received this e-mail in error please immediately notify our Helpdesk on +44 (0) 20 8207 9555. This footnote also confirms that this email message has been swept by MIMEsweeper for the presence of computer viruses. www.mimesweeper.com **********************************************************************
James Ogley said:
My other two machines at home don't experience this at all - one is my firewall, and so doesn't have findutils-locate installed. The other is running 7.2, so although it does have updatedb, it's a different kernel (2.4.4 as opposed to 2.4.18 on 8.0) so it seems to be something that changed between those kernel versions, was it then that the VM system was changed?
It was changed officially in 2.4.10
Or was it changed in the -aa tree in that time (cos Hubert's kernels are based on the -aa tree IIRC)
My memory fails me. Wasn't the -aa tree created for the specific purpose of running Andrea Arcangeli's VM? I honestly don't know. //Anders
It was changed officially in 2.4.10 My memory fails me. Wasn't the -aa tree created for the specific purpose of running Andrea Arcangeli's VM? I honestly don't know.
That's what I thought, but if the official VM was changed to Andrea's, then there'd be no need to persist with it. I just recall someone from SuSE recently saying (it may have been Hubert himself actually - not sure) that Hubert's builds are based on the -aa tree, not the kernel.org tree. 'Course it could have been the -ac tree, it's only a 1 character difference ;) -- James Ogley, Unix Systems Administrator, Pinnacle Insurance Plc james.ogley@pinnacle.co.uk www.pinnacle.co.uk +44 (0) 20 8731 3619 Using Free Software since 1994, running GNU/Linux (SuSE 8.0) This email was created and sent with Ximian Evolution 1.0.8 NEW: Advogato diary at www.advogato.org/person/riggwelter *********************************************************************** CONFIDENTIALITY. This e-mail and any attachments are confidential and may also be privileged. If you are not the named recipient, please notify the sender immediately and do not disclose the contents to another person, use it for any purpose, or store or copy the information in any medium. Any views expressed in this message are those of the individual sender, except where the sender specifically states them to be the views of Pinnacle Insurance Plc. If you have received this e-mail in error please immediately notify our Helpdesk on +44 (0) 20 8207 9555. This footnote also confirms that this email message has been swept by MIMEsweeper for the presence of computer viruses. www.mimesweeper.com **********************************************************************
James Ogley wrote: | | I have to reboot the machine in question every couple of days (I | actually now have a cron job to do it each morning shortly after 5am) | cos otherwise after a couple of days it does start swapping, and it | slows down quite horribly. | I find this a bit puzzling... I'm also running SuSE8.0 with find-utils and all, updatedb in cron, and I see no memory leaks. Strange. I am using a custom compilation of the standard SuSE8 kernel (which is 2.4.18) but I can't think of anything that I've changed that should effect things. I'm interested to see how this turns out. jalal -- GPG fingerprint = 3D45 5509 D380 26A4 523E A9D8 A66A 5F38 CA43 BB0E -- GPG fingerprint = 3D45 5509 D380 26A4 523E A9D8 A66A 5F38 CA43 BB0E
On Thu, Sep 26, 2002 at 10:32:43PM +0200, jalal wrote:
James Ogley wrote: | | I have to reboot the machine in question every couple of days (I | actually now have a cron job to do it each morning shortly after 5am) | cos otherwise after a couple of days it does start swapping, and it | slows down quite horribly. |
I find this a bit puzzling... I'm also running SuSE8.0 with find-utils and all, updatedb in cron, and I see no memory leaks. Strange.
I don't see this problem either and that is really strange. One system: pike@fizia:~> uptime 2:03pm up 76 days, 16:59, 7 users, load average: 0.61, 0.30, 0.27 pike@fizia:~> free total used free shared buffers cached Mem: 256412 253164 3248 0 54068 38816 -/+ buffers/cache: 160280 96132 Swap: 789720 7924 781796 pike@fizia:~> rpm -q k_deflt k_deflt-2.4.18-118 pike@fizia:~> ls /etc/cron.daily/ clean_catman logrotate suse.de-clean-tmp updatedb clean_core suse.de-backup-rc.config suse.de-clean-vi do_mandb suse.de-backup-rpmdb suse.de-cron-local faxcron suse.de-check-battery tetex It's SuSE 8.0. The other system is SuSE 7.3 on sparc. kastus@bursa:~> uptime 2:06pm up 19 days, 20:12, 2 users, load average: 0.50, 0.15, 0.04 kastus@bursa:~> free total used free shared buffers cached Mem: 125304 116240 9064 0 2296 57688 -/+ buffers/cache: 56256 69048 Swap: 130392 24704 105688 kastus@bursa:~> rpm -q k_deflt k_deflt-2.4.14-1 kastus@bursa:~> ls /etc/cron.daily/ aaa_base aaa_base_clean_core aaa_base_rotate_logs aaa_base_backup_rc.config aaa_base_clean_instlog aaa_base_updatedb aaa_base_backup_rpmdb aaa_base_clean_tmp aaa_base_clean_catman aaa_base_do_mandb
I am using a custom compilation of the standard SuSE8 kernel (which is 2.4.18) but I can't think of anything that I've changed that should effect things.
I'm using SuSE stock kernels.
I'm interested to see how this turns out.
I am interested too. Regards, -Kastus
participants (4)
-
Anders Johansson
-
jalal
-
James Ogley
-
Konstantin (Kastus) Shchuka