On Thu, 06 Mar 2014, Kurt Weiske wrote to All:
I've had sporadic issues with ALLFIX version 6.0.24 -- it'll work
for a couple of weeks, then it'll hang, with nothing in the log. Is
there anything I can do to debug this?
what OS?
i know that v6 has had some changes and fixes added to it but the only thing i can relate to is what was happening on my system with my 5.13.4 universal... that being that if one of my frontdoor nodes was rescanning the netmail folder at the same time that allfix was, allfix would lockup if it ran across the same
message that the FD node was on... in programmer speak, this was likely to be because allfix was not opening the messages in compatibility mode... that meaning that it was trying to open the message in a deny mode where no other tools could access the message... i don't remember when i finally figured this out but when i finally did figure out that it was allfix, i moved all my TIC file processing to be done during midnight maintence when the mailer nodes were
all down...
later on, i had to upgrade from my reliable OS/2 Warp 3 to eCS and allfix was locking up again... in some cases, completely locking the entire system... now,
mind you, even though i have the universal flavor of allfix, i was running it completely on the DOS side of my setup... on a whim, i tried running it from the OS/2 side and it worked without locking up! there have been times when it or my OS/2 binkd have locked up and it has turned out to be a conflict of some sort to do with the EMX libraries that both were compiled with... i figured this out because i could close one or the other of the two and the other one would take off finishing its processing... that points to a race condition... probably trying to acquire some semaphore or pipe... i don't know... what i did
find out was that i could run another OS/2 flavor of binkd and avoid the problem... that being the klibc flavor... both binkd/2s are named the same because no one on the binkd team has seen fit to name the klibc one differently
in the make files and it is very easy for me to figure out when i have updated to the wrong flavor because it or allfix will jam up on me... change the flavor
of binkd and we're back to running properly...
i should also note that my setup is geared to the frontdoor layout and operating methods... my binkd stuff mainly uses fileboxes which are managed by my inbound and outbound scripts (4DOS/4OS2 BAT/BTM/CMD files)... allfix (the main one) has no clue about binkd being on the system... i say "the main one" because i do have a secondary instance strictly for handling FREQs that come in
via binkd... it is an exact copy of the main allfix with a couple of paths and file names changed as well as the mailer setting being altered to whatever BSO allfix knows about... it works great and i only need to update the FIX and IDX files in the bink allfix every once in a while... that mainly after adding a new node or making any files areas changes... SETUP.FIX and SETUP.IDX are not copied over when i do this update, though... that would mean that i have to go back in and adjust the paths, file names and mailer settings again as well as reenabling the FREQ processor...
i know this may not help you with your situation but it may help you in finding
out if the lockups you are having are comping from the netmail area processing if your mailer is also scanning the netmail area at the same time... at least, IME, that's the first place i would look...
oh yeah, yes, the OS/2 side of my allfix universal doesn't have the same netmail area scanning problem as the DOS side... this is understandable considering that the universal flavor has separate internal binaries for each OS flavor allfix runs on... they're all just packaged in one large binary with a special jump table at the beginning of the executable that determines which flavor needs to be used ;)
)\/(ark
* Origin: (1:3634/12)