Subject: Re: Connection reset by peer, spawns huge garbage file
From: Benjamin Gilbert (i479@prism.simplenet.com)
Date: Fri May 12 2000 - 18:44:48 EDT
At 5/12/00 7:55 AM, Lancaster, David Matthew (c4ng2@unb.ca) wrote:
>System:
>RH 6.0 Kernel 2.2.5-15
>netatalk1.4b2+asun2.1.3
>
>Problem: Seems like Mac OS 8.5.5 has a problem copying to the box.
>Connection is dropped very quickly, but child afpd process continues
>running, and copys large amounts of garbage data to the file (killed the
>process when it hit 2.1 gigs on a file that was supposed to be 25 megs).
>Syslog mentions numberous (hundreds of thousand) error messages:
>[...]
>May 12 09:30:00 tnt afpd[4098]: dsi_stream_read(-1): Connection reset by
>peer
>May 12 09:30:00 tnt afpd[4098]: dsi_stream_read(0): Connection reset by
>peer
>May 12 09:30:31 tnt last message repeated 15339 times
>[...]
>Somebody mentioned a known bug that spontaneously creates huge files, is
>this it? Any fixes/workarounds? pre2.1.4 the only fix? And if so, which
>version 2.1.4-37b, 2.1.4-38b_test, 2.1.4-39_test?
Don't know about the connection being dropped, but the creation of a
garbage file once a connection dies is the known bug you referred to. No
workarounds short of fixing the code, which -- practically speaking --
means upgrading to pre2.1.4. pre2.1.4 is stable; I've been running it on
a production server for 8 months and I haven't had any problems.
You'd probably want to go with 39_test unless you need codepage support,
in which case you should use 37b; the _test means that asun has codepage
support disabled while he reworks it.
--Benjamin Gilbert
This archive was generated by hypermail 2b28 : Wed Jan 17 2001 - 14:30:43 EST