home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!snorkelwacker.mit.edu!ai-lab!dead!ead.dsa.com!schorr
- From: schorr@ead.dsa.com (Andrew J. Schorr)
- Newsgroups: comp.sys.sun.admin
- Subject: Re: Running out of mbufs
- Keywords: SunOS mbuf
- Message-ID: <1992Sep4.112756@ead.dsa.com>
- Date: 4 Sep 92 15:27:56 GMT
- Sender: news@ead.dsa.com
- Organization: Daiwa Securities America, New York
- Lines: 59
-
- Last week I posted a message concerning a SparcStation 1
- running SunOS 4.1.2 that repeatedly hangs (every 10 days
- or so) because it runs out of mbufs.
-
- It gives error messages before it freezes that look like this:
- Aug 25 04:13:06 ead2 vmunix: mbuf map full
- Aug 26 02:47:51 ead2 vmunix: le0: out of mbufs - packet dropped
- Aug 26 02:54:24 ead2 last message repeated 18 times
- Aug 26 02:57:16 ead2 vmunix: le0: out of mbufs - packet dropped
- Aug 26 02:57:25 ead2 vmunix: le0: out of mbufs - packet dropped
-
- A number of people have responded, but no one seems to have
- a definitive solution.
-
- Halvard Halvorsen suggested increasing the MAXUSERS parameter
- (presumably this will cause the kernel to allocate more mbufs).
-
- Brad Albom mentioned a problem he's having that relates
- to the number of mbufs allocated to data steadily
- increasing over time. (I don't think this is my problem,
- since this number seems to fluctuate up and down for me.)
-
- rjq@phys.ksu.edu (Rob Quinn) suggested looking for relevant patches.
- a906270@tiuk.ti.com (Simon Willis) noted that patch 100330 for
- SunOS 4.1.1 talks about mbuf problems and claims that they
- are obsoleted by 4.1.2 plus patches 100516 and 100570.
- However, when you look
- at the README's for those 2 patches, they don't seem to be related
- to this problem. I may install them anyway.
-
- brock@cs.unca.edu mentioned that his Ultrix machines have the
- same problem.
-
- Well, that's it. I may increase MAXUSERS and/or install patches
- 100516 and 100570, but I'm not confident.
-
- In case anyone can figure out some new ideas, here are the results
- of a netstat -m performed shortly before my most recent hang:
-
- 419/512 mbufs in use:
- 81 mbufs allocated to data
- 7 mbufs allocated to packet headers
- 137 mbufs allocated to socket structures
- 173 mbufs allocated to protocol control blocks
- 3 mbufs allocated to routing table entries
- 13 mbufs allocated to socket names and addresses
- 3 mbufs allocated to zombie process information
- 2 mbufs allocated to interface addresses
- 1980/1980 cluster buffers in use
- 2044 Kbytes allocated to network (99% in use)
- 3028 requests for memory denied
- 0 requests for memory delayed
- 0 calls to protocol drain routines
-
- Once again, if you have any further thoughts, please respond
- by e-mail to schorr@ead.dsa.com, since my news feed is down.
-
- Thanks,
- Andy
-