Number: 115

Date: 17-Mar-84  0':18':38

Submitter: Sannella.PA

Source: desRivieres, dering.pasa (CitiBank), Roach, MGittins.RX

Subject: NS filing reliability problems

Assigned To: vanMelle

Attn: Release

Status: Fixed

In/By: Harmony

Problem Type: Performance

Impact: Serious

Difficulty: Hard

Frequency: Intermittent

Priority: Absolutely

System: Communications

Subsystem: NS Protocols



Lisp Version: 

Source Files: 

Microcode Version: 

Memory Size: 

File Server: 

Server Software Version: 

Disposition: '
[Date':  9 Apr 84 12':38 PST'
AR 115': Hard.]'
["Masinter" "18-Sep-84 21':05':41" Source': Assigned% To': Attn': Status':(Open->Fixed)]'
["Sannella.PA" "19-Sep-84 18':53':53" In/By':]'
["Sannella.PA" "19-Sep-84 18':54':12" In/By': Disposition': Description': Edit-By': Edit-Date':]

Description: '
Date': 16 Feb 84 19':25 PST'
Subject': Lisp': NS Filing': Raised Priority'
To':, Cooper, Sheil'
Lisp-System-Date': 14-Feb-84 13':57':51'
Machine-Type': Dolphin'
There seems to be a general concensus here at PARC that the Lisp-NSFiling problems are limited to lack of random access.  Au contraire, there are plenty of Lisp problems lurking there.  Perhaps we have been lulled by a lack of fresh reports on the problem.'
Syntelligence have had so many problems NSFiling that they have basically quit using it, and are relying on floppy and flaky {dsk} to store their stuff.  They have however been able to sysout to their fileserver successfully with their pre-Carol lisp (thanks largely to Bill''s work on this problem).  Makefile etc. are apparently still a disaster.'
NSFiling was a considerable nuisance in the lispcourse in Watford.   File interactions usually (though by no means always) completed successfully.  But they frequently seemed to spawn 3-5 filing processes each, n-1 of which would get confused and periodically promptprint "Filing#i not responding" ad infinitum.  Why it takes more than one process to accomplish a makefile I don''t know.  It seems as if all the spurious processes were waiting for the fileserver to tell them something, but were not reminding the fileserver they were waiting to be served.  And sometimes file interactions broke, usually when trying to close a file. People at the course who had NS fileservers at home said that they had the same problems on their own installations. '
It seems that after the thaw, we must allocate some resources to fixing the Lisp end of NSFiling.  For starters, we must appoint some Phylex users': to the best of my knowledge, NO ONE is using it now.  I for one volunteer.  But I do not wish to be alone.  And once NSFiling seems solid, I don''t see any reason why we can''t tackle transparently caching files on the local disk -- it can''t take more than a couple days to design and implement.'
-- Tayloe.'
Date': 21 Feb. 1984 1':18 pm PST (Tuesday)'
From': Raim.pasa'
Subject': Re': Lisp': NS Filing': Raised Priority'
In-reply-to':''s message of 16 Feb 84 19':25 PST'
I am considering Jerry Fung as the NS-networking resource for this problem.  It'
would require his spending some time at PARC though.  If no one has a better'
suggestion, perhaps you can take this up with beau.'
Date':  5 MAR 84 16':16 PST'
From': ROACH.PA'
Subject': PHYLEX': BUGS'
cc':   ROACH'
     (1) I don''t get prompted by PHYLUM or MAXC for password.  I think '
it is unreasonable that PHYLEX': demands my password.'
     (2) Subdirectory creation should happen automatically rather'
than PHYLEX':''s current "FILE NOT FOUND" breakage.'
     (3) I don''t think "XSIS North':Xerox" should be part of PHYLEX':''s'
device name.  I also feel the same way about "LispPrint':XSIS North".'
     (4) It''s pretty unpleasant to see an interminable number of'
"PHYLEX': Filing not responding"''s appear in my promptwindow when'
I am not even interacting with the beastie.'
Date':  5 Mar 84 19':26 PST'
Subject': Lisp': NS File servers'
cc': 3LispSupport↑.pa'
Lisp-System-Date':  5-Mar-84 16':21':20'
Machine-Type': Dandelion'
Some of the problems that I have been having with Phylex':'
1.  (fullname ''foo ''new) always returns nil; however, (fullname ''foo ''old) seems fine.'
2.  Tedit PUT to a non-existent file causes a FILE NOT FOUND in OPENSTREAM.  If the file exists, a different error is generated.'
3.  Tedit get on an empty file causes a break under SPP.GETBYTE.'
4.  Assorted processes named "PHYLEX': Filing#xx" buzz around like mosquitos printing "PHYLEX': Filing#xx not responding.".   They have to be swatted with PSW KILL.'
5.  Doing a MAKESYS to Phylex': caused "SPP Retransmit Queue out of order" error.'
Date': 17 Apr 84 00':07 PST'
Subject': Lisp': NS Filing gnats '
Lisp System Date':  8-Apr-84 18':02':02'
Machine': Dorado (DewDrop)'
Microcode version': 24,4'
Memory size': 10000'
Frequency': Always'
Impact': Moderate'
Question': what purpose in life do the NS host filing processes serve.'
There seems to be too many of them, and they seem to loiter forever, (erroneously?) reporting "NSHost': Filing not responding." in the prompt window.   This tends to give the user the impression that the file server is on the blink, which it usually isn''t. '
To see what I mean, I think all you need do is a COPYFILE to an NS host, and then wait a couple of minutes.'
---Jim '
[AR 1112 from Ed Szemis -- Citibank  (Dering.pasa)'
  18-May-84 11':26':38]'
                     Citibank is in the process of moving all of their files from an IFS to NS file server.  They are unable to compile files on the NS file server that successfully compiled on the IFS.  They get an error in the courier program filing with the following backtrace':'
Date': 18 Sep 84 21':05 PDT'
Subject': AR#115, NS Filing reliability problems'
To': LispSupport'
cc': vanMelle, 1100Support.pasa, Sheil'
In Bill''s absence, I am declaring AR#115, which is the "grandaddy" AR of NS Filing problems, as Fixed. I imagine there are still a number of minor glitches in NS Filing, and that ARs on them will arise, but all of the major show-stoppers (can''t compile, extra processes, odd messages, poor response, etc.) all seem to work, at least in the cases that I''ve tried.'
I''d wait for Bill, but I''d like to see this AR# in the list of "fixed" ARs.'


Test Case: 

Edit-By: Sannella.PA

Edit-Date: 19-Sep-84 18':54':13