martelli@cadlab.sublink.ORG (Alex Martelli) (02/27/91)
Help! Our ATP's are way down, our line keeps falling... what's happening to us? This was supposed to be a hardware UPgrade...:-) Until last week we used, as uucp connection to the world, an old and trusty HP9000/340 and a T1000; we used to have ATP's in the 700's with HP/UX 6.2, then (a pleasant surprise!) in the 800's after upgrading to HP/UX 7.0 (no other changes). The builtin serial async port of the HP340 appeared to lack any modem control, so we used it in "direct" mode, i.e. a minor device of 0x900004. The T1000 was accordingly set for xon/xoff flow control, with the interface locked at 9600. The port, and the modem, were (and are) ONLY used for uucp: indeed, there is NO incoming line, just an OUTgoing one (through a PABX, but that had never given any problem) for this sole purpose. Last week we upgraded to an HP 9000/400t, a much more powerful machine (lots of RAM, lots of disk, faster clock, etc), and HP/UX 7.03B. All SOFTware (B-news, elm, nn, etc) appears to work admirably... BUT the ATP on the line have dropped horribly, and the line keeps failing completely. The old machine and the new one are still side by side; the physical change was just unplugging the modem from the 340 and plugging it into the serial port of the 400, with the same straight through flat cable (yes, we DID check - it's SECURELY connected!-). Everything from the cable to the modem to the phone line is the same. Reconnecting the 340 "solves" all problems; unfortunately this is NOT a viable option in the long term, since the 340's destined to go away reasonably soon. We first tried continuing to use a minor of 0x900004, as had worked so admirably on the 340; ATP disaster, for days (see later for detailed symptoms). Today we went at it hammer and tongs and tried everything: cua00/cul00/tty00 triples with both 'ccitt' and 'simple' modem control, with all possibile flow settings on the T1000 (rts/cts, xon/xoff, both, neither); locking the interface at both 9600 and 19200; nothing good happened. What else can we try??? HELP!...please... Some details: according to data from /usr/spool/uucp/.Admin/xferstats, ATPs vary between 250 and 350 when we lock the interface at 9600; they go between 400 and 450 when we lock it at 19200 instead. Observation at 9600 shows a pattern: reception of a file starts with a "burst" - RD on solidly, TD wildly flashing (in acknowledgment?) for several seconds; then several seconds of pause; now another "burst" as above, but meanwhile uucico, if at suitably high debug level, is writing "alarm N" - the "alarm" message appears maybe half a second AFTER the "burst" starts arriving; all file arrives in similar "bursts", and EACH "burst" appears to "trigger" (?) an "alarm" message. At 19200 each "burst" seems to become made up of three portions: a pre-burst (RD *solid* red, TD flashing); a mid-burst, shorter (RD *pulsing* red, TD flashing); an after-burst (RD *solid* red again, TD flashing); only the pre-burst is accompanied by the alarm. Eventually carrier drops and uucico fails with "OO short 0 want 3" message. Luckily our feeder node is making LOTS of connect time available to us for this emergency, still we have BIG trouble raking up all of our data, particularly large files (our feeder's Cnews sometimes seems to make bundles of 150-200K for us, although it's SUPPOSED to cut it up in 50k-pieces) - our connection is rarely up long enough to gather a 160k-file at one gulp, for example (when it DOES make it, it's about 500+ seconds at 9600, 400 at 19200). It seems there's occasional loss of characters if I use this setup to dial my home machine (an Interactive-Unix 386 box with T1000) with ct. Never happened before on similar ct's when the machine on this side was the hp340 box. Any help will be greatly appreciated, as we're *totally* at a loss. Thanks in advance! -- Alex Martelli - CAD.LAB s.p.a., v. Stalingrado 53, Bologna, Italia Email: (work:) martelli@cadlab.sublink.org, (home:) alex@am.sublink.org Phone: (work:) ++39 (51) 371099, (home:) ++39 (51) 250434; Fax: ++39 (51) 366964 (work only), Fidonet: 332/401.3 (home only).
rjn@hpfcso.FC.HP.COM (Bob Niland) (03/03/91)
re: > Our ATP's are way down, our line keeps falling... What's an ATP? > The builtin serial async port of the HP340 appeared to lack any modem > control, so we used it in "direct" mode, i.e. a minor device of 0x900004. It has TxD,RxD,CD,RI,DSR,DTR,CTS,RTS > The T1000 was accordingly set for xon/xoff flow control, with the > interface locked at 9600. I'm surprised that this worked at all. The 340 has no buffering on the serial port. I would expect problems above 4800. > Last week we upgraded to an HP 9000/400t, a much more powerful machine > (lots of RAM, lots of disk, faster clock, etc), and HP/UX 7.03B. The 400 Models have a 16 byte FIFO and CTS/RTS hardware flow control available on the serial port at select code 9 (as do the 345 and 375). You need 7.05 or a patch to enable these features. If your local HP support representative does not have the patch, have them contact me. Regards, Hewlett-Packard Bob Niland Internet: rjn@FC.HP.COM 3404 East Harmony Road UUCP: [hplabs|hpfcse]!hpfcrjn!rjn Ft Collins CO 80525-9599