--
--PART-BOUNDARY=.19506021518.ZM25329.cup.hp.com
Content-Type: text/plain; charset=us-ascii
Hi All,
Here is the last (please) proposal for ":listf,access". I have tried to
incorporate many of your good suggestions into this draft. This proposal
only describes :listf,8 == :listfile,access. I will send you another proposal
for :listfile,9 == :listfile,locks which will provide more details to help
solve possbile file deadlock situations.
If you see any serious flaws please let me know asap.
Much thanks,
Jeff Vance
--------------------
Format 8 ("access"):
--------------------
:listfile logfil@, access
********************
LOGFILE1.LOGGING1.SYSTEM12
0 Accessors
********************
LOGFILE2.LOGGING1.SYSTEM12
0 Accessors
********************
LOGFILE3.LOGGING1.SYSTEM12
5 Accessors(O:5,R:3,W:2,L:4),Shared,Remote
#S12345 JOBNAME8,USER5678.ACCT5678 R:2,W:2,L:3 LDEV: 12345
#S40 FRED,MGR.PAYROLL R:1,W:0,L:1 REM: 15.127.45.120
#J15 JSPOOL,RSPOOL.SYS R:1,W:0,L:0 SPID: #O7654321
:listfile logfil@, 8; seleq=[accessed=TRUE]
********************
LOGFILE3.LOGGING1.SYSTEM12
5 Accessors(O:5,R:3,W:2,L:4),Shared,Remote
#S12345 JOBNAME8,USER5678.ACCT5678 R:2,W:2,L:3 LDEV: 12345
#S40 FRED,MGR.PAYROLL R:1,W:0,L:1 REM: 15.127.45.120
#J15 JSPOOL,RSPOOL.SYS R:1,W:0,L:0 SPID: #O7654321
:listfile @.work,8;seleq=[accessed=TRUE]
********************
OCIMAIN.WORK.ACCT
1 Accessor(O:0,R:0,W:1,L:1),Shared,Store
#J15 JDAILY,BACKUP.SYS R:1,W:0,L:1 SPID: #O76
As in all existing HFS formats, the data starts in col. 2 rather than col 1.
:listfile /bin/a@,8
********************
/bin/awk
2 Accessors(O:2,R:2,W:0,L:1),Shared
#S12 CATHY.DEVELOP R:1,W:0,L:0 LDEV: 228
#J8 JCOMP,MBETOOLS.BUILD R:1,W:0,L:1 SPID: #O369
:listfile /Spec@;format=access;tree
********************
/Special/
0 Accessors
********************
/Special/Long_file_name_directory/
0 Accessors
********************
/Special/Long_file_name_directory/this_is_a_ridiculous_name/
0 Accessors
********************
/Special/Long_file_name_directory/this_is_a_ridiculous_name/lets-wrap_and_make_
life_miserable/
0 Accessors
********************
/Special/Long_file_name_directory/this_is_a_ridiculous_name/lets-wrap_and_make_
life_miserable/foo
1 Accessor(O:1,R:0,W:1,L:0),Exclusive
#S55 MGR.PAYROLL R:0,W:1,L:0 LDEV: 541
Note: the directories shown above are listed because ;tree was specified;
however, these directories are not being accessed even though "foo" is.
A :listfile /,8;notree will show all logged on users, since all logons
formally open root.
A :listfile /SYS,8 will show all users logged onto the SYS account,
since all logons formally open their logon account.
A :listfile /SYS/,8;seleq=[object=dir] will show all users logged onto
the SYS account and all users who have a CWD at any level below SYS
(including logon groups in SYS).
Note: ";seleq=[accessed=true]" would have eliminated all of the parent
directory lines above.
Accessor Summary Line:
----------------------
The accessor line (under the filename) shows the total number of accessors
across all job/sessions for the file. This total includes all processes that
have formally opened the file and all processes that have pseudo opened the
file (sm_open). This is the same number seen in the "FLAGS" field of :listf,3.
The "O:" field is the actual number of processes that formally opened the file
(e.g., called FOPEN, HPFOPEN, open, etc.). For those that care, this is how
many PLFD entries exist for this file. What if the O: number is different from
the total number? This does happen (but O: is never > total) and means that
one or more processes are keeping this file available until they complete there
work. One example of this is when the file is being STOREd. Also, :listf,2
will sm_open file to collect data. If you are doing a :listf,8 on the same
file that someone else is doing a :listf,2 you will see the total count higher
than the O: count. The total number of readers, writers and lockers (L:) are
shown next.
If the file is being shared then ",Shared" is displayed; otherwise ",Exclusive"
is displayed, in which case, there is typically only a single accessor
(exception is POSIX fork and dup). If the file is being stored then one of
",Storing", ",Restoring" or ",ConcurrBackup" is shown. If any accessor is
remotely connected then ",Remote" is shown. If everything above is true the
longest accessor line is:
nnnnn Accessors(O:nnnnn,R:nnnn,W:nnnn,L:nnnn),Exclusive,ConcurrBackup,Remote
The #J / #S Data Line:
----------------------
The job/session id is displayed followed by the optional jobname, followed
by the user.account names. The group name is not seen. Next the number
of processes (in that job/session) that are reading, writing and locking the
file is displayed. The last field depends on job vs. session vs. remote
session. For all jobs "#spoolfile name" (in .OUT.HPSPOOL) is shown after the
"SPID:" label. For all local (host-originating) sessions the ldev is shown.
For all remote sessions "REM:" is displayed followed by the IP address.
Here's how a remote session is determined:
if the HPVTPIN CI variable exists (thanks Eero!) then we have an incoming VT
connection and that VTSERVER's PIN is used to get the IP addr of the socket
else if the accessor's PIN has an incoming socket open we use it's IP addr
{ => dscopy (NFT.net.sys or RASERVER.net.sys), ftp (FTPSRVR.arpa.sys),
or home grown }
else if the accessor's PIN's JSMAIN has an incoming socket open we use it's
IP addr { => telnet server }
else { assume local CI, get ldev }
Here are maxed out accessor lines:
#S12345 JOBNAME8,USER5678.ACCT5678 R:8286208,W:8286208,L*LDEV: 12345
#S12345 JOBNAME8,USER5678.ACCT5678 R:1024,W:8286208,L* REM: xxx.xxx.xxx.xxx
#S12345 JOBNAME8,USER5678.ACCT5678 R:0,W:8286208,L* REM: xxx.xxx.xxx.xxx
#S12345 JOBNAME8,USER5678.ACCT5678 R:0,W:123456,L:8286208LDEV: 12345
Note: there is always room for the max number of readers+writers; however,
there is not always enough room for the max R+W+lockers. In this case,
"L*" indicates loss of data.
Questions for format=ACCESS
----------------------------------------------------------------------------
#1 In the last example (/Special/....) where there is only 1 accessor
the "O:1,R:0,W:1,L:0" is redundant. Should the summary accessor line,
immediately below the filename line, be dropped when there is only
one accessor?
----------------------------------------------------------------------------
#2 For ;seleq=[ACCESSED=TRUE]: there are two kinds of accessors, 1) a
normal open of the file, 2) a special open of the file that is used
to prevent the file (or related data structures) from disappearing
(sm_open). This second open does not return a file descriptor (fnum)
and is used by backup products, :listf, and many other tools. I assume
that you consider an accessor to be both 1) and 2), but please let me
know if this is true, or is a bad idea. Also, some files are protected
by the system by being opened at boot time (PROGEN is the accessor).
Do you consider these files as being accessed, and do you want to see
PROGEN (pin 1) as the accessor?
----------------------------------------------------------------------------
#3 Do I need to worry about enough room for IPng addresses?
----------------------------------------------------------------------------
Please let me know what you think. Thanks for all of your great input!
Regards,
Jeff Vance, MPE Lab
--
Jeff Vance
--PART-BOUNDARY=.19506021518.ZM25329.cup.hp.com
Content-Description: Transferred from mail from Mail Delivery Subsystem <[log in to unmask]>: Session Transcript
Content-Type: text/plain ; charset=us-ascii ; charset=us-ascii
550 3000-L... User unknown
--PART-BOUNDARY=.19506021518.ZM25329.cup.hp.com
Content-Description: Transferred from mail from Mail Delivery Subsystem <[log in to unmask]>: Returned Content
Content-Type: message/rfc822
Received: by hpmpes3.cup.hp.com
(1.37.109.11/15.5+IOS 3.20+cup+OMrelay) id AA253231350; Fri, 2 Jun 1995 15:15:51 -0700
Return-Path: <[log in to unmask]>
From: "Jeff Vance" <[log in to unmask]>
Message-Id: <[log in to unmask]>
Date: Fri, 2 Jun 1995 15:15:49 -0700
X-Mailer: Z-Mail (3.2.0 06sep94)
To: [log in to unmask]
Subject: 3rd (and last!?) :listf,access draft
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Hi All,
Here is the last (please) proposal for ":listf,access". I have tried to
incorporate many of your good suggestions into this draft. This proposal
only describes :listf,8 == :listfile,access. I will send you another proposal
for :listfile,9 == :listfile,locks which will provide more details to help
solve possbile file deadlock situations.
If you see any serious flaws please let me know asap.
Much thanks,
Jeff Vance
--------------------
Format 8 ("access"):
--------------------
:listfile logfil@, access
********************
LOGFILE1.LOGGING1.SYSTEM12
0 Accessors
********************
LOGFILE2.LOGGING1.SYSTEM12
0 Accessors
********************
LOGFILE3.LOGGING1.SYSTEM12
5 Accessors(O:5,R:3,W:2,L:4),Shared,Remote
#S12345 JOBNAME8,USER5678.ACCT5678 R:2,W:2,L:3 LDEV: 12345
#S40 FRED,MGR.PAYROLL R:1,W:0,L:1 REM: 15.127.45.120
#J15 JSPOOL,RSPOOL.SYS R:1,W:0,L:0 SPID: #O7654321
:listfile logfil@, 8; seleq=[accessed=TRUE]
********************
LOGFILE3.LOGGING1.SYSTEM12
5 Accessors(O:5,R:3,W:2,L:4),Shared,Remote
#S12345 JOBNAME8,USER5678.ACCT5678 R:2,W:2,L:3 LDEV: 12345
#S40 FRED,MGR.PAYROLL R:1,W:0,L:1 REM: 15.127.45.120
#J15 JSPOOL,RSPOOL.SYS R:1,W:0,L:0 SPID: #O7654321
:listfile @.work,8;seleq=[accessed=TRUE]
********************
OCIMAIN.WORK.ACCT
1 Accessor(O:0,R:0,W:1,L:1),Shared,Store
#J15 JDAILY,BACKUP.SYS R:1,W:0,L:1 SPID: #O76
As in all existing HFS formats, the data starts in col. 2 rather than col 1.
:listfile /bin/a@,8
********************
/bin/awk
2 Accessors(O:2,R:2,W:0,L:1),Shared
#S12 CATHY.DEVELOP R:1,W:0,L:0 LDEV: 228
#J8 JCOMP,MBETOOLS.BUILD R:1,W:0,L:1 SPID: #O369
:listfile /Spec@;format=access;tree
********************
/Special/
0 Accessors
********************
/Special/Long_file_name_directory/
0 Accessors
********************
/Special/Long_file_name_directory/this_is_a_ridiculous_name/
0 Accessors
********************
/Special/Long_file_name_directory/this_is_a_ridiculous_name/lets-wrap_and_make_
life_miserable/
0 Accessors
********************
/Special/Long_file_name_directory/this_is_a_ridiculous_name/lets-wrap_and_make_
life_miserable/foo
1 Accessor(O:1,R:0,W:1,L:0),Exclusive
#S55 MGR.PAYROLL R:0,W:1,L:0 LDEV: 541
Note: the directories shown above are listed because ;tree was specified;
however, these directories are not being accessed even though "foo" is.
A :listfile /,8;notree will show all logged on users, since all logons
formally open root.
A :listfile /SYS,8 will show all users logged onto the SYS account,
since all logons formally open their logon account.
A :listfile /SYS/,8;seleq=[object=dir] will show all users logged onto
the SYS account and all users who have a CWD at any level below SYS
(including logon groups in SYS).
Note: ";seleq=[accessed=true]" would have eliminated all of the parent
directory lines above.
Accessor Summary Line:
----------------------
The accessor line (under the filename) shows the total number of accessors
across all job/sessions for the file. This total includes all processes that
have formally opened the file and all processes that have pseudo opened the
file (sm_open). This is the same number seen in the "FLAGS" field of :listf,3.
The "O:" field is the actual number of processes that formally opened the file
(e.g., called FOPEN, HPFOPEN, open, etc.). For those that care, this is how
many PLFD entries exist for this file. What if the O: number is different from
the total number? This does happen (but O: is never > total) and means that
one or more processes are keeping this file available until they complete there
work. One example of this is when the file is being STOREd. Also, :listf,2
will sm_open file to collect data. If you are doing a :listf,8 on the same
file that someone else is doing a :listf,2 you will see the total count higher
than the O: count. The total number of readers, writers and lockers (L:) are
shown next.
If the file is being shared then ",Shared" is displayed; otherwise ",Exclusive"
is displayed, in which case, there is typically only a single accessor
(exception is POSIX fork and dup). If the file is being stored then one of
",Storing", ",Restoring" or ",ConcurrBackup" is shown. If any accessor is
remotely connected then ",Remote" is shown. If everything above is true the
longest accessor line is:
nnnnn Accessors(O:nnnnn,R:nnnn,W:nnnn,L:nnnn),Exclusive,ConcurrBackup,Remote
The #J / #S Data Line:
----------------------
The job/session id is displayed followed by the optional jobname, followed
by the user.account names. The group name is not seen. Next the number
of processes (in that job/session) that are reading, writing and locking the
file is displayed. The last field depends on job vs. session vs. remote
session. For all jobs "#spoolfile name" (in .OUT.HPSPOOL) is shown after the
"SPID:" label. For all local (host-originating) sessions the ldev is shown.
For all remote sessions "REM:" is displayed followed by the IP address.
Here's how a remote session is determined:
if the HPVTPIN CI variable exists (thanks Eero!) then we have an incoming VT
connection and that VTSERVER's PIN is used to get the IP addr of the socket
else if the accessor's PIN has an incoming socket open we use it's IP addr
{ => dscopy (NFT.net.sys or RASERVER.net.sys), ftp (FTPSRVR.arpa.sys),
or home grown }
else if the accessor's PIN's JSMAIN has an incoming socket open we use it's
IP addr { => telnet server }
else { assume local CI, get ldev }
Here are maxed out accessor lines:
#S12345 JOBNAME8,USER5678.ACCT5678 R:8286208,W:8286208,L*LDEV: 12345
#S12345 JOBNAME8,USER5678.ACCT5678 R:1024,W:8286208,L* REM: xxx.xxx.xxx.xxx
#S12345 JOBNAME8,USER5678.ACCT5678 R:0,W:8286208,L* REM: xxx.xxx.xxx.xxx
#S12345 JOBNAME8,USER5678.ACCT5678 R:0,W:123456,L:8286208LDEV: 12345
Note: there is always room for the max number of readers+writers; however,
there is not always enough room for the max R+W+lockers. In this case,
"L*" indicates loss of data.
Questions for format=ACCESS
----------------------------------------------------------------------------
#1 In the last example (/Special/....) where there is only 1 accessor
the "O:1,R:0,W:1,L:0" is redundant. Should the summary accessor line,
immediately below the filename line, be dropped when there is only
one accessor?
----------------------------------------------------------------------------
#2 For ;seleq=[ACCESSED=TRUE]: there are two kinds of accessors, 1) a
normal open of the file, 2) a special open of the file that is used
to prevent the file (or related data structures) from disappearing
(sm_open). This second open does not return a file descriptor (fnum)
and is used by backup products, :listf, and many other tools. I assume
that you consider an accessor to be both 1) and 2), but please let me
know if this is true, or is a bad idea. Also, some files are protected
by the system by being opened at boot time (PROGEN is the accessor).
Do you consider these files as being accessed, and do you want to see
PROGEN (pin 1) as the accessor?
----------------------------------------------------------------------------
#3 Do I need to worry about enough room for IPng addresses?
----------------------------------------------------------------------------
Please let me know what you think. Thanks for all of your great input!
Regards,
Jeff Vance, MPE Lab
--
Jeff Vance
--PART-BOUNDARY=.19506021518.ZM25329.cup.hp.com--
|
