During the Solaris installation/upgrade, the logs will be written into
/var/sadm/system/logs/<install/upgrade>_log

Are there any logs(for Solaris) written during the normal run-time
i.e. after the Solaris installation/upgrade is completed?

Basically, I want to have some health checks(for the complete software
stack) for the system where my application is running. Hence, I am
trying to find such health check steps for checking Solaris OS itself.

qazmlp wrote:

> During the Solaris installation/upgrade, the logs will be written into
> /var/sadm/system/logs/<install/upgrade>_log
> 
> Are there any logs(for Solaris) written during the normal run-time
> i.e. after the Solaris installation/upgrade is completed?
>
No, unfortunately Solaris doesn't create any system logs whatsoever.  It 
makes administration of this operating system a complete nightmare, 
because when something goes wrong, there's absolutely no way you can 
troubleshoot it.  If you encounter any problems while running Solaris, 
the only solution is to completely reinstall it.

On the plus side however Solaris Engineers generally get paid a lot more 
than those working with other operating systems.

> Basically, I want to have some health checks(for the complete software
> stack) for the system where my application is running. Hence, I am
> trying to find such health check steps for checking Solaris OS itself.

Hmm, I think your best bet is to run:

truss -p 1 -o /tmp/system.log &

And whenever you want to see how your operating system is performing, 
just run:

tail -f /tmp/system.log

In article <9q5bd.1080316$yk.181980@news.easynews.com>,
 Mothra <mothra@mothra.com> wrote:

> qazmlp wrote:
> 
> > During the Solaris installation/upgrade, the logs will be written into
> > /var/sadm/system/logs/<install/upgrade>_log
> > 
> > Are there any logs(for Solaris) written during the normal run-time
> > i.e. after the Solaris installation/upgrade is completed?
> >
> No, unfortunately Solaris doesn't create any system logs whatsoever.  It 
> makes administration of this operating system a complete nightmare, 
> because when something goes wrong, there's absolutely no way you can 
> troubleshoot it.  If you encounter any problems while running Solaris, 
> the only solution is to completely reinstall it.
> 
> On the plus side however Solaris Engineers generally get paid a lot more 
> than those working with other operating systems.
> 
> > Basically, I want to have some health checks(for the complete software
> > stack) for the system where my application is running. Hence, I am
> > trying to find such health check steps for checking Solaris OS itself.
> 
> Hmm, I think your best bet is to run:
> 
> truss -p 1 -o /tmp/system.log &
> 
> And whenever you want to see how your operating system is performing, 
> just run:
> 
> tail -f /tmp/system.log

"PAY NO ATTENTION TO THAT MAN BEHIND THE CURTAIN."

I have no idea if Mothra (and her sidekick Rodan) were jerking your 
chain or just channeling a deceased M$ support tech.  

In truth, Solaris 8 for the Ultra processors (sun4u architecture) have 
prtdiag.  It depends on what hardware you're running but you can search 
for it at http://docs.sun.com/?q=prtdiag to check out your hardware.  
Older hardware (e.g. sun4m--Sparc 20 and that era) doesn't have this 
diagnostic for temperature, system boards, I/O boards, etc because it 
was simpler.

Also, extensive logging occurs from the kernel when there's memory or 
other problem on boot.  Try

dmsg

or

view /var/adm/messages

-- 
DeeDee, don't press that button!  DeeDee!  NO!  Dee...

"Michael Vilain <vilain@spamcop.net>" wrote:

> 
> 
> "PAY NO ATTENTION TO THAT MAN BEHIND THE CURTAIN."
> 
> I have no idea if Mothra (and her sidekick Rodan) were jerking your 
> chain or just channeling a deceased M$ support tech.  
>
That must have been the Evil Mothra, my arch nemesis.  He/she/it has a 
cruel and cynical sense of humour indeed.

Nice Mothra says:  "All your Solaris operating system logs are generated 
by /usr/sbin/syslogd which reads in the configuration file 
/etc/syslog.conf (and stores its PID in /etc/syslog.pid).

I wouldn't recommend editing /etc/syslog.conf unless you really know 
what you're doing - the default settings should be fine for novices and 
the syntax is extremely tricky to get right.

But if you ignore that warning, BE VERY CAREFUL when editing 
/etc/syslog.conf, there must be no space characters - only tab 
characters.  I've lost count of the number of Solaris Systems I've 
encountered where syslogd is running along merrily, but nothing is being 
written to /var/adm/messages, only to find that someone's idly cut 'n' 
pasted /etc/syslog.conf from another machine and all of the tab 
characters have been converted to spaces.

With regard to diagnostic tools on Solaris, there are so many it's 
difficult to know where to begin, but here are some commands to get you 
started:

vmstat - to check your virtual memory
iostat - to check your I/O activity, e.g. disk drives
netstat - to check your network settings

You could also cheat and download "top" from www.sunfreeware.com, but 
don't ever admit this to experienced Solaris Admins, as they will laugh 
derisorily at you (especially that Evil Mothra guy).

For monitoring your application, you should check out the so-called 
"proc tools" (under /usr/proc/bin).  The man pages for these commands 
will show you how you can view all the threads of a running program.

And don't forget "truss" - my all-time favourite last resort debugging 
tool.  Check out this article:
http://sunsolve.sun.com/pub-cgi/show.pl?target=content/content10

Also, you might want to try the wonderful SE Tools for more advanced 
diagnostics.

And make sure you browse through http://www.sun.com/bigadmin/ for other 
gems of information.

But if you really want ninja skills in this department, send yourself on 
SA-400 Solaris System Performance Management (or its equivalent if the 
name's changed again).  I have yet to attend a better Sun Microsystems 
training course.

In comp.unix.solaris Mothra <mothra@mothra.com> wrote:
> No, unfortunately Solaris doesn't create any system logs whatsoever.  It 
> makes administration of this operating system a complete nightmare, 

Whatever you're smoking, I want some...

  Scott

Scott Howard wrote:

> In comp.unix.solaris Mothra <mothra@mothra.com> wrote:
> 
>>No, unfortunately Solaris doesn't create any system logs whatsoever.  It 
>>makes administration of this operating system a complete nightmare, 
> 
> 
> Whatever you're smoking, I want some...
> 
>   Scott

/usr/bin/hash ?

(sorry)

Mothra <mothra@mothra.com> writes:

>
>> Basically, I want to have some health checks(for the complete software
>> stack) for the system where my application is running. Hence, I am
>> trying to find such health check steps for checking Solaris OS itself.
>
> Hmm, I think your best bet is to run:
>
> truss -p 1 -o /tmp/system.log &
>
> And whenever you want to see how your operating system is performing, just
> run:
>
> tail -f /tmp/system.log

Nice. Will have to remember that...

Dragan

-- 
Dragan Cvetkovic, 

To be or not to be is true. G. Boole      No it isn't.  L. E. J. Brouwer

!!! Sender/From address is bogus. Use reply-to one !!!

Dragan Cvetkovic wrote:
> Mothra <mothra@mothra.com> writes:
> 
> 
>>>Basically, I want to have some health checks(for the complete software
>>>stack) for the system where my application is running. Hence, I am
>>>trying to find such health check steps for checking Solaris OS itself.
>>
>>Hmm, I think your best bet is to run:
>>
>>truss -p 1 -o /tmp/system.log &
>>
>>And whenever you want to see how your operating system is performing, just
>>run:
>>
>>tail -f /tmp/system.log
> 
> 
> Nice. Will have to remember that...
> 
> Dragan
> 

Dragan, unless I am completely bonkers, trussing init will give you no 
information whatsoever about "how your operating system is performing". 
My truss just tells me when init is rereading /etc/inittab for the most 
part. The rest is pause()...

Beardy <beardy@beardy.net> writes:

> Dragan Cvetkovic wrote:
>> Mothra <mothra@mothra.com> writes:
>>
>>>>Basically, I want to have some health checks(for the complete software
>>>>stack) for the system where my application is running. Hence, I am
>>>>trying to find such health check steps for checking Solaris OS itself.
>>>
>>>Hmm, I think your best bet is to run:
>>>
>>>truss -p 1 -o /tmp/system.log &
>>>
>>>And whenever you want to see how your operating system is performing, just
>>>run:
>>>
>>>tail -f /tmp/system.log
>> Nice. Will have to remember that...
>> Dragan
>>
>
> Dragan, unless I am completely bonkers, trussing init will give you no
> information whatsoever about "how your operating system is performing". My
> truss just tells me when init is rereading /etc/inittab for the most
> part. The rest is pause()...
>

You are right. I suppose trussing pid 0 would be more interesting, but in
that case I am getting

bash-2.05b# truss -f -p 0 
truss: system process: 0

Ditto for e.g. 2

Shame.

Bye, Dragan

-- 
Dragan Cvetkovic, 

To be or not to be is true. G. Boole      No it isn't.  L. E. J. Brouwer

!!! Sender/From address is bogus. Use reply-to one !!!

Mothra <mothra@mothra.com> writes:

> You could also cheat and download "top" from www.sunfreeware.com, but
> don't ever admit this to experienced Solaris Admins, as they will
> laugh derisorily at you (especially that Evil Mothra guy).

I use top myself, but when I come across people who don't know what it
tells them, I do laugh derisorily, I have to admit :-

"I think the machine is busy, look at this top output"

"is that % figure per cpu or total"

"oh, err, I don't know"

"which version of top is this"

"not sure"

"is it the same as on this other window on the other server"

"yes ... oh no, look at that, no it's completely different"

<oh for heaven's sake>"Just use prstat -a, ok?"

Chris
-- 
Chris Morgan
   "Post posting of policy changes by the boss will result in 
    real rule revisions that are irreversible"

		- anonymous correspondent

Beardy <beardy@beardy.net> writes:

> Dragan, there is rat all that you can do with system processes as they are
> not processes as we know of. In fact, I suspect that there is no real need
> for ps to list them. I s'pose it is just to keep us happy that there is
> *something* using PID 0. Why pageout is not PID 1 I don't know - prob just
> historical.

I guess so too. But then everybody expects init to be pid 1 as much as
everybody expect root to be uid 0.

Dragan

-- 
Dragan Cvetkovic, 

To be or not to be is true. G. Boole      No it isn't.  L. E. J. Brouwer

!!! Sender/From address is bogus. Use reply-to one !!!

Dragan Cvetkovic wrote:
> Beardy <beardy@beardy.net> writes:
> 
> 
>>Dragan Cvetkovic wrote:
>>
>>>Mothra <mothra@mothra.com> writes:
>>>
>>>
>>>>>Basically, I want to have some health checks(for the complete software
>>>>>stack) for the system where my application is running. Hence, I am
>>>>>trying to find such health check steps for checking Solaris OS itself.
>>>>
>>>>Hmm, I think your best bet is to run:
>>>>
>>>>truss -p 1 -o /tmp/system.log &
>>>>
>>>>And whenever you want to see how your operating system is performing, just
>>>>run:
>>>>
>>>>tail -f /tmp/system.log
>>>
>>>Nice. Will have to remember that...
>>>Dragan
>>>
>>
>>Dragan, unless I am completely bonkers, trussing init will give you no
>>information whatsoever about "how your operating system is performing". My
>>truss just tells me when init is rereading /etc/inittab for the most
>>part. The rest is pause()...
>>
> 
> 
> You are right. I suppose trussing pid 0 would be more interesting, but in
> that case I am getting
> 
> bash-2.05b# truss -f -p 0 
> truss: system process: 0
> 
> Ditto for e.g. 2
> 
> Shame.
> 
> Bye, Dragan
> 

Dragan, there is rat all that you can do with system processes as they 
are not processes as we know of. In fact, I suspect that there is no 
real need for ps to list them. I s'pose it is just to keep us happy that 
there is *something* using PID 0. Why pageout is not PID 1 I don't know 
- prob just historical.

Beardy.

In article <lmd5zm1prg.fsf@privacy.net>, Dragan Cvetkovic wrote:
> Beardy <beardy@beardy.net> writes:
> 
>> Dragan, there is rat all that you can do with system processes as they are
>> not processes as we know of. In fact, I suspect that there is no real need
>> for ps to list them. I s'pose it is just to keep us happy that there is
>> *something* using PID 0. Why pageout is not PID 1 I don't know - prob just
>> historical.
> 
> I guess so too. But then everybody expects init to be pid 1 as much as
> everybody expect root to be uid 0.

Uhhh, root HAS to be UID 0 - its the UID that is special, not the name 'root'...

Kevin

spamtotrash@toomuchfiction.com (Kevin Collins) writes:

> In article <lmd5zm1prg.fsf@privacy.net>, Dragan Cvetkovic wrote:
 
>> I guess so too. But then everybody expects init to be pid 1 as much as
>> everybody expect root to be uid 0.
>
> Uhhh, root HAS to be UID 0 - its the UID that is special, not the name 'root'...

I know. Let me rephrase myself: why root is uid 0 and not e.g. 314159 (or
42) is probably just historical.

The same way that O_RDONLY is defined as 0 in <sys/fcntl.h> confusing the
hell out of (new) programmers who do open(file, O_RDONLY|O_WRONLY) and
wonder why is it different than open(file, O_RDWR) ...

Or why is it creat(2) and not create(2) (see
http://web.umr.edu/~eepe/jon.html#Thompson for the context).

Bye, Dragan

-- 
Dragan Cvetkovic, 

To be or not to be is true. G. Boole      No it isn't.  L. E. J. Brouwer

!!! Sender/From address is bogus. Use reply-to one !!!

In article <lm4qky1m9t.fsf@privacy.net>, Dragan Cvetkovic wrote:
> spamtotrash@toomuchfiction.com (Kevin Collins) writes:
> 
>> In article <lmd5zm1prg.fsf@privacy.net>, Dragan Cvetkovic wrote:
>  
>>> I guess so too. But then everybody expects init to be pid 1 as much as
>>> everybody expect root to be uid 0.
>>
>> Uhhh, root HAS to be UID 0 - its the UID that is special, not the name 'root'...
> 
> I know. Let me rephrase myself: why root is uid 0 and not e.g. 314159 (or
> 42) is probably just historical.
> 
> The same way that O_RDONLY is defined as 0 in <sys/fcntl.h> confusing the
> hell out of (new) programmers who do open(file, O_RDONLY|O_WRONLY) and
> wonder why is it different than open(file, O_RDWR) ...

Well, everything in UNIX is historical - it wasn't created all at once out of
the ether :) It was created, at some time other than right now, and hence is
historical by definition!

> Or why is it creat(2) and not create(2) (see
> http://web.umr.edu/~eepe/jon.html#Thompson for the context).

Nice link - some good quotes there.

Kevin

Mothra wrote:
> Scott Howard wrote:
> 
>> In comp.unix.solaris Mothra <mothra@mothra.com> wrote:
>>
>>> No, unfortunately Solaris doesn't create any system logs whatsoever.  
>>> It makes administration of this operating system a complete nightmare, 
>>
>>
>>
>> Whatever you're smoking, I want some...
>>
>>   Scott
> 
> 
> /usr/bin/hash ?
> 
> (sorry)

Oh, the pains are coming back .....

In article <MI6bd.1085240$yk.183036@news.easynews.com>, Mothra wrote:

>I wouldn't recommend editing /etc/syslog.conf unless you really know 
>what you're doing - the default settings should be fine for novices and 
>the syntax is extremely tricky to get right.

I think that's untrue.  It's only about as hard as inetd.conf .

ntp.conf, httpd.conf, smb.conf and sudoers are all harder aren't they ?

-- 
Elvis Notargiacomo  master AT barefaced DOT cheek
http://www.notatla.org.uk/goen/
    7.031: OnACPower returned value( 0x1 ) which is Equal To 0x1

Dragan Cvetkovic <me@privacy.net> writes:
> Mothra <mothra@mothra.com> writes:
> > tail -f /tmp/system.log
> Nice. Will have to remember that...

Note that  less +F /tmp/system.log  will do the same.
A nice feature is that you can stop it using ^C and
use pattern highlighting (e.g. -i/error) and continue
again with command F or navigate with < > b f n N etc.


Markus

Dragan Cvetkovic wrote:
> spamtotrash@toomuchfiction.com (Kevin Collins) writes:
> 
> 
>>In article <lmd5zm1prg.fsf@privacy.net>, Dragan Cvetkovic wrote:
> 
>  
> 
>>>I guess so too. But then everybody expects init to be pid 1 as much as
>>>everybody expect root to be uid 0.
>>
>>Uhhh, root HAS to be UID 0 - its the UID that is special, not the name 'root'...
> 
> 
> I know. Let me rephrase myself: why root is uid 0 and not e.g. 314159 (or
> 42) is probably just historical.

314159?

Now I remember... Sharp APL!

Definitely historical :-)

> The same way that O_RDONLY is defined as 0 in <sys/fcntl.h> confusing the
> hell out of (new) programmers who do open(file, O_RDONLY|O_WRONLY) and
> wonder why is it different than open(file, O_RDWR) ...
> 
> Or why is it creat(2) and not create(2) (see
> http://web.umr.edu/~eepe/jon.html#Thompson for the context).
> 
> Bye, Dragan
>