View Issue Details
ID | Project | Category | View Status | Date Submitted | Last Update |
---|---|---|---|---|---|
0000090 | Squid for OS/2 | public | 2006-09-22 10:14 | 2020-08-24 03:14 | |
Reporter | BruceDudley | Assigned To | psmedley | ||
Priority | normal | Severity | crash | Reproducibility | always |
Status | closed | Resolution | fixed | ||
Summary | 0000090: Squid Cache version 2.6.STABLE3 | ||||
Description | always afer about a 24 hour period Squid will exit the the CMD shell. | ||||
Additional Information | In the log file the first error in the log: is: " comm_select: select failure: (14) Bad address examine_select: Examining open file descriptors..." Then following are some 30000+ entries in the log file with the error: "WARNING: FD 106 has handlers, but it's invalid. FD 106 is a None called '' tmout:0 read:0 write:0 FD 109: (9) Bad file number" The number following the FD is incremented up to 32755. Hence the large amount of log entries. There are no log entries at the point of time that Squid exists to the CMS shell indicating what the error was. | ||||
Tags | No tags attached. | ||||
|
A bit more useful info on the problem Killed by SIGSEGV pid=0x0132 ppid=0x011e tid=0x0001 slot=0x00c7 pri=0x0200 mc=0x0001 C:\SQUID\SBIN\SQUID.EXE LIBC061 0:00008fee cs:eip=005b:1df48fee ss:esp=0053:001ccb58 ebp=001ccb88 ds=0053 es=0053 fs=150b gs=0000 efl=00012216 eax=00000000 ebx=00000000 ecx=00000041 edx=00000104 edi=000023ff esi=000023ff Process dumping was disabled, use DUMPPROC / PROCDUMP to enable it. |
|
Can you see if http://smedley.info/squid-2.6s4-os2-b1.zip helps? squid doesn't crash here.... |
|
Can you see if http://smedley.info/squid-2.6s4-os2-b1.zip [^] helps? Thanks Paul, I have tried it out and I still get the problem. However one thing that is interesting. I have two WSEB boxes, one is use on a bussines and it's running 24/7 and so is Squid [your version]. Now these two OS/2 boxes [mine and the other box]are for all intents and purposes, identical software wise. Their OS/2 box does not crash mine does! The _only_ diffrence is that I use CMD.EXE as the default command processor where-as the other os/2 box uses 4os2.exe. I chaned my config.sys to use 4os2.exe as the default command processor and bingo, no more Squid crash. I have no idea why it make the diffrence but it does. So far my Squid's been up for two days and not crashed. Hope this may help and not hinder (: |
|
Sadly that makes no sense at all. I use cmd.exe here and no issues. Can you share your squid.conf so I can see if there any differences from mine that may explain the issues? |
|
I had actually tried using the copy of the squid.conf from the other OS/2 box and apart from the cached directory, it was without any other changes. Still woud not work. Also included the cache.log file with the errors mentioned. You can grab my conf from here: ftp://guest:dunhill@ftp.niclan.homeip.net/squid.zip Thanks Paul. |
|
Hi guys! Bruce - on the box that does not crash, is there a setting for EMXOPTS (which I realise might not be used by the Innotek GCC) in the CONFIG.SYS? Is it the same as the stable machine? Also, are you running any other GCC software on the box that crashes? (e.g. BIND/NAMED, Apache?) Lastly, are both boxes using the same filesystem for the cache directory (and/or log directories)? There may be issues with Squid and earlier builds of JFS... need a new'un for it to be stable. Regards, Stephen |
|
"is there a setting for EMXOPTS " Yes there is. Both OS/2 boxes are virtually identicial as fas as installed OS and software. I have even gone to the extreme of putting my OS/2 HDD into another computer all-together. IE. diffrent mother board + power suplpy. Still the same. Both computers [the one that does not crash with Squid and my one that does] have JFS and both have the same version/build. In my computer I have tried the squid cache to be on HPFS and JFS with either SCSI or IDE HDD's. It still crashes. Both computers uses Pauls builds of Bind9, MySQL v5 and PHP. This _is_ really odd. Why on one it's fine, the other no good! One thing that _maybe_ has an affect is the disk cache of the the JFS/HPFS386. I usually have a 64meg cache and if I'm lucky Squid will stay up for a bit longer than 24hrs, bit if I make it a 32meg cache, Squid will stay up for a lot longer; perhaps 3 days. I just do not know. Sigh! |
|
Hi Bruce, Sometimes HPFS386 can get its knickers in a twist. If you are using HPFS386, please consider giving this a try: Edit IBM386FS\HPFS386.INI and add MAXHEAP = 20480 (assuming you have enough spare memory in your box). If you have less than 256Mb of RAM, you might want to lower MAXHEAP to 8192. But the default algorithum for working out MAXHEAP, if not set, is poor. Or... however, the problem could be to do with spare memory in your system (the 32Mb cache system may have more spare memory than the 64Mb system) - use PMPatrol or Theseus to find out how much real RAM you have spare. When Squid rebuilds its cache indexes, it can use 300Mb.... Regards, Stephen |
|
Some more useless info on my Squid issue. Now after 8 days of 24/7 use, Squid has not crashed on me. Although I do not know what specific thing I changed to rectify to problem. What I have done is to 'start' Squid before LAN server starts up, use a 'call' rather than a 'run' in the config.sys file for cache386.exe, deleted all ACL in the HPFS386 file system. just use NET.ACC instead. No other changes [that I can remember] Squid is now stable. Given that both the main Squid directory and cached directory are on a JFS HDD's, I know, nothing makes sense but ther you go. Furtermore, as I have said I have another OS/2 box that was a mirror image of my home HDD install and without any fiddling or ado, Squid and the therefore the computer have been up without issue for now 30+ days. Go figure. |
|
Seems to be more of an OS problem than a Squid bug - glad you found a workaround that seems to be working! |
|
The problem is back after a reboot. First reboot in 8 days. 2006/10/20 16:15:29| NETDB state saved; 0 entries, 0 msec 2006/10/20 17:33:50| NETDB state saved; 0 entries, 47 msec 2006/10/20 17:36:39| comm_select: select failure: (14) Bad address 2006/10/20 17:36:39| examine_select: Examining open file descriptors... 2006/10/20 17:36:39| FD 96: (9) Bad file number 2006/10/20 17:36:39| WARNING: FD 96 has handlers, but it's invalid. 2006/10/20 17:36:39| FD 96 is a None called '' 2006/10/20 17:36:39| tmout:0 read:0 write:0 2006/10/20 17:36:39| FD 97: (9) Bad file number 2006/10/20 17:36:39| WARNING: FD 97 has handlers, but it's invalid. I then get hundreds of these entries in the cache.log. My error log is now 3meg |
|
Hi Bruce, If you want, please give http://users.anytimenow.com/will_arrive/squid-2.6.s4-os2-beta.zip a go. It contains the EXE files only, so you will need to backup your existing Squid directory and then replace the EXE files in it. I can't claim all of the credit for this build, but it seems to be stable although I have not tested it under production load. It is built with Pinger support and X-Forwarded For support enabled. Logfile rotate and redirector support should also now be working, but I'm waiting for some updated SquidGuard code from Paul before I can finish testing. I'd be very interested if you could post back here any feedback or results... Regards, Stephen |
|
Hi, The uploaded Squid file at http://users.anytimenow.com/will_arrive/squid-2.6.s4-os2-beta.zip has been updated because of a rather large hot sugar+butter issue, so if you have already downloaded it, then you will need to get it again, sorry. An issue was the with code for enabling x-forwarded-for, which I've backed out for now. The vesion on test has Delay Pool, HTCP, NULL storage and Heap processing available. Sorry for the fudge - when this is all done I'll send the code back to Paul to see if it will build with the latest LibC. Regards, Stephen |
|
[ The uploaded Squid file a ] Yup, many thanks. I have got it and running so far without issue. [ I'd be very interested if you could post back here any feedback or results ] Having said that, one thing that crops up, going to "hobbes.nmsu.edu/" only displays a blank page. I need to clear the Firefox [win32] & IE cache each time before it will display. Another site this occurs with is 'www.bankwest.com.au". Plus one or two other. It occurs at random, that is, sometimes I visit the site and all is OK, next time, [next day] it's a blank page. If you want me to try any other releases, I be happy to oblige. |
|
Hi Bruce, (and Paul!) Have a go if you want with a new build at http://users.anytimenow.com/will_arrive/squid-2.6.s4-os2-beta6.zip Previous builds have persistent connection support broken (which caused long realplayer streams to abort, SSL connections to fry and blank/invalid downloads). This build (Stable 4 at 2006-10-31) includes HTCP, Delay Pool, ICMP and redirector/rotate support with all known problems fixed - with the exception of the 'blank page' problem which I can't say for certain is fixed until it has been tested more. Still won't compile and build a reliable EXE with Netlab's Libc, gone back to EMX... ;-( The LIBC builds still come up with the select_failure problem and the 32,000 log entries. Henrik at squid-cache.org very kindly built me a pure POSIX compliant version of the Squid I/O loop in the last week, and even *this* won't work under Libc (which I guess is still a work in progress...) Lastly, this build also *requires* TCP/IP v4, as this build for performance enhancement directly uses a certain few IP functions directly from the OS/2 v4 IP stack, rather than use EMX. Please let me have any feedback you may have. I'll write this all up for Paul at some stage - he's really the master at all this kind of stuff! Hopefully we might also be able to give some useful feedback to the Libc developer, who is working very hard on it... Regards, Stephen |
|
Hi guys, I got exactly the same problems as Bruce, starting from version 3.0.stable6. I reverted to 3.0.stable4 which runs fine (can't remember if I ever tried 3.0.5). I am getting confused by the number you use: is this about an older version? Piersante |
|
3.0.5 does the same. stable4 is the last that works here Piersante |
|
Please open a new bug if this is still present in 3.0.stable26 |
Date Modified | Username | Field | Change |
---|---|---|---|
2006-09-22 10:14 | BruceDudley | New Issue | |
2006-09-23 07:37 | BruceDudley | Note Added: 0000298 | |
2006-09-24 09:43 | psmedley | Note Added: 0000304 | |
2006-09-26 11:34 | BruceDudley | Note Added: 0000307 | |
2006-09-26 11:35 | BruceDudley | Note Edited: 0000307 | |
2006-09-26 11:36 | BruceDudley | Note Edited: 0000307 | |
2006-09-26 12:04 | psmedley | Note Added: 0000308 | |
2006-09-26 13:10 | BruceDudley | Note Added: 0000309 | |
2006-09-26 13:52 | BruceDudley | Note Edited: 0000309 | |
2006-09-30 02:36 | Stephen | Note Added: 0000310 | |
2006-10-03 14:25 | BruceDudley | Note Added: 0000313 | |
2006-10-03 14:26 | BruceDudley | Note Edited: 0000313 | |
2006-10-03 14:27 | BruceDudley | Note Edited: 0000313 | |
2006-10-04 19:56 | Stephen | Note Added: 0000314 | |
2006-10-16 03:28 | psmedley | Project | Other Unix Port => Squid for OS/2 |
2006-10-19 12:07 | BruceDudley | Note Added: 0000359 | |
2006-10-19 12:15 | psmedley | Note Added: 0000360 | |
2006-10-19 12:15 | psmedley | Assigned To | => psmedley |
2006-10-19 12:15 | psmedley | Status | new => feedback |
2006-10-21 02:30 | BruceDudley | Note Added: 0000364 | |
2006-10-21 02:33 | BruceDudley | Note Edited: 0000364 | |
2006-10-21 02:52 | BruceDudley | Note Edited: 0000364 | |
2006-10-21 12:35 | Stephen | Note Added: 0000367 | |
2006-10-22 17:00 | Stephen | Note Added: 0000368 | |
2006-10-26 10:56 | BruceDudley | Note Added: 0000398 | |
2006-10-26 10:58 | BruceDudley | Note Edited: 0000398 | |
2006-11-01 03:43 | Stephen | Note Added: 0000422 | |
2008-11-17 18:16 | piesse | Note Added: 0001022 | |
2008-12-11 09:53 | piesse | Note Added: 0001077 | |
2011-12-31 08:25 | psmedley | Note Added: 0002104 | |
2011-12-31 08:25 | psmedley | Status | feedback => resolved |
2011-12-31 08:25 | psmedley | Resolution | open => fixed |
2020-08-24 03:14 | psmedley | Status | resolved => closed |