setsite in tcl

[Todd Horst]

Can you setsite in tcl, without using the exec command.

If i use exec, once it completes, Im back to the site i started off in.

This is in the context of a tcl script running at the system level, and not inside an interface

Thanks

Topic

[Keith McLeod]

Not sure if this will work for you.

hcisetenv -site

I ran this from a hcitcl prompt.

[Jeff Dinsmore]

I banged my head against this one for a while…

The solution I settled on is to reset env variables in the Tcl environment.

Here’s a proc to do that.  It calls Cloverleaf’s hcisetenv (a perl script), then uses the return from that to set environment variables in Tcl.

It’s not the most elegant, but seems to work well enough.

Code:

namespace eval cloverUtils {

}

proc cloverUtils::setSite { site } {

global env

set envSet [exec [file join $env(HCIROOT) bin hcisetenv] -site $site]

for { set i 0 } { $i < [llength $envSet] } { incr i } {
switch -exact — [lindex $envSet $i] {
set {
set n [expr $i + 1]
set v [expr $i + 2]
set env([lindex $envSet $n]) [regsub -all {'} [lindex $envSet $v] ""]
incr i 2
}
clear {
incr i
}
}
}

 return 1

}

Jeff Dinsmore
Chesapeake Regional Healthcare

[David Barr]

You can also do something like this:

Code:

set site test_adt
global env
catch { exec $::HciRoot/sbin/hcisetenv -site tcl $site } vars
eval “$vars”

But this won’t do a few other things that the command line “setsite”  command does like setting HciSite, HciSiteDir, auto_path, etc.

[Jeff Dinsmore]

David the issue with your approach is that, while it does set local variables in the Tcl shell you’re running, it doesn’t set environment variables in the Tcl shell.

That works fine as long as you only want to use the variable values locally, but if you want to exec a command from Tcl in the scope of a site other than the one your Tcl shell started in, you need to change the env() array variables in your current shell.

Jeff Dinsmore
Chesapeake Regional Healthcare

[David Barr]

Actually my approach does update the env array.

[Russ Ross]

Instead of struggling with changing sites inside TCL, I went to doing it in the K-Shell that wraps around my TCL.

Here is a URL to illustrate that approach

<a href="http://clovertech.infor.com/viewtopic.php?t=6876&#8243; class=”bbcode_url”>http://clovertech.infor.com/viewtopic.php?t=6876

Russ Ross
RussRoss318@gmail.com

[Jeff Dinsmore]

Russ –

Todd’s question was about setting site from inside Tcl.

I may have missed something, but the post in the link you supplied looks to address only KSH.

Jeff Dinsmore
Chesapeake Regional Healthcare

[Russ Ross]

You are correct about my URL showing an approach of using KSH instead of TCL to change sites.

I was attempting to illustrate another way to approach looping thru sites that worked for me that was less hassle as opposed to doing it in TCL.

I wasn’t even sure it was possible to loop thru sites in TCL correctly until msiDetach came along, and by then I already had an apporach of using the KSH site change wrapper.

So basically the KSH script changes sites effectively without observed issues or side effects and calls the TCL script(s) that assume the site has already been set when it is called.

If staying with TCL 100% is one of your conditions, then my URL posted appoach will not meet that condition.

Russ Ross
RussRoss318@gmail.com

[Jeff Dinsmore]

David –

I’ll be dipped – so it does…  Looks like I worked too hard at that one.

I didn’t notice the “tcl” flag – didn’t know it existed.

I had not found the option.  It’s not listed in the  usage at the shell level and it’s buried in my CL documentation.

So, Todd.  I’d do it David’s way ;0)

Jeff Dinsmore
Chesapeake Regional Healthcare

[Jeff Dinsmore]

Russ –

You’re absolutely right.  If KSH is your starting point, there’s no pressing need to do it all in Tcl – unless you want to.

My need – and I’m assuming Todd’s as well – is primarily from within Tcl UPOC/TPS code inside of Cloverleaf.  From there, it’s much cleaner if you can stay in Tcl.

Jeff Dinsmore
Chesapeake Regional Healthcare

[Todd Horst]

Thanks for everyones reply, and sorry for the late response.

Im trying to do something like:

Code:

global env
eval [exec ksh -c “$env(HCIROOT)/sbin/hcisetenv -site tcl site1″]
msiAttach
msiGetStatSample Thread1
msiDetach

eval [exec ksh -c “$env(HCIROOT)/sbin/hcisetenv -site tcl site2″]
msiAttach
msiGetStatSample Thread2
msiDetach

This will work for the first thread, outputing the stats correctly. However when i try to do the msiAttach on the next site i get the following error:

Code:

tcl>msiAttach
[0:TEST] oilSysSemOpen: ftok failed
Sem Name = /NetConfig Project Id : 90
[0:TEST] msiGlobalInit: Can’t open guard sem: No such file or directory
msiGlobalInit: NetConfig = /NetConfig
Error: shared memory attach failed

Has anyone run into this before?

[Richard Hart]

Hi Todd.

We are on *ix and generally run one script that loops through the Cloverleaf sites and then calls a second script to perform  the action.

When the second script has completed its work, the envieronment is dropped until the next call.  The second script can also be called from  the command line.

eg

cycleAllLog.sh calling

 cycleLog.sh

cycleAllSmat.sh calling

cycleSmat.sh

purgeAllSite.sh calling

 purgeSite.sh

[David Barr]

Todd Horst wrote:

This will work for the first thread, outputing the stats correctly. However when i try to do the msiAttach on the next site i get the following error:

Code:

tcl>msiAttach
[0:TEST] oilSysSemOpen: ftok failed
Sem Name = /NetConfig Project Id : 90
[0:TEST] msiGlobalInit: Can’t open guard sem: No such file or directory
msiGlobalInit: NetConfig = /NetConfig
Error: shared memory attach failed

Has anyone run into this before?

It seems like I’ve run into this and I don’t remember what I did to fix it.

If you want to stay 100% tcl, one possible solution is to loop through all of your sites, call “fork” to spawn a child process. The child process could call msiAttach/msiDetach/exit. The parent process could call “wait” to wait for the child process to finish with each site.

[Todd Horst]

i was thinking something along this line… ill give it a go

[Russ Ross]

Even though my KSH wrapper for changing sites that calls TCL scripts takes care of my needs, this topic of using all TCL became a splinter in my brain.

I tried all kinds of variations inside a single TCL proc to loop thru sites but always ran into trouble when doing the msiAttach past the first site just like the rest of you indicated.

I asked Steve Fraser and he had the same outcome but did make me aware of a built in cloverleaf TCL proc named ChangeSite that looks promising.

Then I started thinking about a TCL wrapper approach since my KSH wrapper approach works and that showed promise as the following proof of concept examples produced the outcome I was looking for.

Naturally I was looking at a small set of outcomes to determine if success was possible, so swim at your own risk.

Here is my /hcitest/russ/test_changing_sites.tcl standalone wrapper script:

Code:

#!/usr/bin/env hcitcl

set siteList [list test_picis test_ris]

#—————————————————————————————————————-
# method 1 proof of concept using built-in Cloverleaf ChangeSite TCL proc that Steve Fraser told me about, thanks
#—————————————————————————————————————-

echo “n########## method 1 proof of concept ##########”
echo [exec showroot]

foreach site $siteList {
   ChangeSite $site
   echo [exec /hcitest/russ/test_msiTocEntry.tcl]
}

#——————————————————————————————-
# method 2 proof of concept using clovertech example from David Barr to eval “$vars”, thanks
#——————————————————————————————-

echo “n########## method 1 proof of concept ##########”
echo [exec showroot]

foreach site $siteList {
   catch { exec $::HciRoot/sbin/hcisetenv -site tcl $site  } vars
   eval “$vars”
   echo [exec /hcitest/russ/test_msiTocEntry.tcl]
}

echo “n########## Done ##########”
echo “n[exec showroot]n”

Here is the /hcitest/russ/test_msiTocEntry.tcl standalone TCL script:

Code:

#!/usr/bin/env hcitcl

echo “n===== test_msiTocEntry.tcl”
echo “HciRoot ($HciRoot)”
echo “HciSite ($HciSite)”
echo “”

msiAttach
echo [msiTocEntry]
msiDetach

I got to wondering if msiDetach was even necessary when using a TCL wrapper approach like this so I commented it out and my test still worked fine with no complaints about aleady being attached.

I find myself wondering If I understand how msiDetach is intended to be used because the next msiAttach after the first msiDetach never worked when inside the same TCL script and it doesn’t appear to be necessary when using a wrapper approach.

Here is the output from running /hcitest/russ/test_changing_sites.tcl scripts

Code:

########## method 1 proof of concept ##########
HCI root is /cloverleaf/cis6.0/integrator
HCI master site is mda_global
HCI site is test_russ

===== test_msiTocEntry.tcl
HciRoot (/cloverleaf/cis6.0/integrator)
HciSite (test_picis)

mock_picis_ormgr_v75_adt ob_picis_bldbank_ordrslts dump_picis_ormgr_v72_adt ob_lab_picis_v71 load_adt_picis_v71 dump_picis_ormgr_v75_adt ob_picis_ormgr_v81_adt ob_picis_lab_dev ob_adt_picis_v71 9216_test_lab ob_adt_picis_newserver rs_picis_ormgr_v72_adt jr_bldbnk_ordrslt load_picis_ormgr_v72_adt rs_picis_ormgr_v75_adt ob_picis_ormgr_v72_adt load_adt_ormgr_v72 ob_picis_ormgr_v75_lab mock_picis_ormgr_v72_adt 9231_test_adt load_picis_ormgr_v75_adt ob_picis_ormgr_v75_adt load_adt_ormgr_v75

===== test_msiTocEntry.tcl
HciRoot (/cloverleaf/cis6.0/integrator)
HciSite (test_ris)

ob_syngo ob_pwrscribe jr_sms_genie_8238 ob_ris_train ib_ris_test_orders hsfw_8239 js_coderyte_10002 hrfw_8331 hrfw_8332 ob_stentor_adt ib_pwrscribe_ORU ib_syngo ib_ris_train_orders hrfw_8239 ob_ris_test ob_magview_oru hsfw_8331 ob_sms_la hsfw_8332

########## method 1 proof of concept ##########
HCI root is /cloverleaf/cis6.0/integrator
HCI master site is mda_global
HCI site is test_ris

===== test_msiTocEntry.tcl
HciRoot (/cloverleaf/cis6.0/integrator)
HciSite (test_picis)

mock_picis_ormgr_v75_adt ob_picis_bldbank_ordrslts dump_picis_ormgr_v72_adt ob_lab_picis_v71 load_adt_picis_v71 dump_picis_ormgr_v75_adt ob_picis_ormgr_v81_adt ob_picis_lab_dev ob_adt_picis_v71 9216_test_lab ob_adt_picis_newserver rs_picis_ormgr_v72_adt jr_bldbnk_ordrslt load_picis_ormgr_v72_adt rs_picis_ormgr_v75_adt ob_picis_ormgr_v72_adt load_adt_ormgr_v72 ob_picis_ormgr_v75_lab mock_picis_ormgr_v72_adt 9231_test_adt load_picis_ormgr_v75_adt ob_picis_ormgr_v75_adt load_adt_ormgr_v75

===== test_msiTocEntry.tcl
HciRoot (/cloverleaf/cis6.0/integrator)
HciSite (test_ris)

ob_syngo ob_pwrscribe jr_sms_genie_8238 ob_ris_train ib_ris_test_orders hsfw_8239 js_coderyte_10002 hrfw_8331 hrfw_8332 ob_stentor_adt ib_pwrscribe_ORU ib_syngo ib_ris_train_orders hrfw_8239 ob_ris_test ob_magview_oru hsfw_8331 ob_sms_la hsfw_8332

########## Done ##########

HCI root is /cloverleaf/cis6.0/integrator
HCI master site is mda_global
HCI site is test_ris

Russ Ross
RussRoss318@gmail.com

[Russ Ross]

With regards to msiAttach, I consider it prudent to check to see if the $HCISIEDIR/exec/monitorShmemFile exist and if it doesn’t then I don’t run msiAttach.

I don’t know about current versions of cloverleaf but in the past running msiAttach would create $HCISIEDIR/exec/monitorShmemFile after which mysterous problems would surface.

This was an effective workaround since if no $HCISIEDIR/exec/monitorShmemFile exist, then the site most likely is not running.

Here is an example:

Code:

if { [lsearch -exact [ls $HciSiteDir/exec] monitorShmemFile] == -1 } { exit }

Russ Ross
RussRoss318@gmail.com

[Bob Richardson]

Greetings,

Running CIS 6.1.0 on AIX 6 TL9 (HA virtualized).

Russ, our Tcl programs that invoke the msiAttach comannd surround the call with a catch.  For a site we find that you can only call msiAttach once in a TCL instance (running program).  Otherwise you get an error but it is not fatal.  Keeps the return values “clean”.

Sample code:

    if {$mode == “start”} {

         catch {msiAttach}

    } elseif {$mode == “run”} {

       set threadstats [msiGetStatSample $url_thread]

       keylget threadstats OBDATAQD pend_count  ;#get the pending count for the requeue thread

Hope this helps.

Enjoy.

[Jeff Dinsmore]

We are on 5.6, so don’t have msiDetach, but do recall attempting to just change sites and then run msiAttach again which just crashed and burned…

So, to loop through all sites for diagnostic/monitoring, I found it necessary to exec a call to hcicmd to get thread/process info.

Code:

set resultList [lrange [exec hcicmd -p hcimonitord -t d -c “statusrpt {$crmcNetConfigArray($site,$process,threadList)}”] 2 end]

I then evaluate the returned resultList and glean whatever info I’m looking for.

Using exec was not my preference, but looked to be the only option for what I was trying to do.

Jeff Dinsmore
Chesapeake Regional Healthcare

[Todd Horst]

Thanks jeff, that looks like the same data, without the messiness of msi.

Russ, what you have works, i kind of wanted it to be one script, or at least have that ability.

I started looking into interp. And made it pretty far, however I still get the same with when doing an msiAttach on a subsequent site.

Code:

set s [interp create]
proc hello2 {} {
puts “hello2”
}

proc mykeylkeys {arg} {
return [keylkeys arg]
}
proc mykeylget {arg1 arg2} {
return [keylget arg1 $arg2]
}

# Procs that they need access to
interp alias $s hello2 {} hello2
interp alias $s msiAttach {} msiAttach
interp alias $s msiDetach {} msiDetach
interp alias $s msiGetStatSample {} msiGetStatSample
interp alias $s keylkeys {} mykeylkeys
interp alias $s keylget {} mykeylget
interp alias $s echo {} echo

# Variables that they need access to
interp eval $s [list “set” “threads” $threads]
interp eval $s [list “set” “auto_path” $::auto_path]
#interp eval $s [list “set” “env” $env]

set threads [interp eval $s {
proc hello {} {
puts “hiya”
}
global env
#hello
#hello2
#puts $auto_path
msiAttach

set idx -1
foreach thread $threads {

# Get name from dictionary
set Name [dict get $thread Name]

# Get stats
set result [msiGetStatSample $Name]

set StatKEYS [keylkeys $result]

# Protocal Status (UP/DOWN/CONNECTING)
set Status “Unknown”
if {[lsearch -exact $StatKEYS “PSTATUS”] != -1} {
set Status [string totitle [keylget $result PSTATUS]]
}
dict append thread Status $Status
}
 
msiDetach
return $threads
}]

interp delete $s

return the first site, but then the seoncd site gets the following error on attach

Code:

[0:TEST] oilSysSemOpen: ftok failedSem Name = /NetConfig Project Id : 70[0:TEST] msiGlobalInit: Can’t open guard sem: No such file or directorymsiGlobalInit: NetConfig = /NetConfig

For now I think ill go with jeffs command

[Russ Ross]

I find myself liking Jeff’s approach, too.

I tested it on a site where the monitod wasn’t running and it returns an error code of 2, plus doesn’t start the monitord or create a shared memory file that can cause problems later on like it does with msiAttach.

Russ Ross
RussRoss318@gmail.com

[Author]

Replies