Re: Solaris OS versions supported on the cool thread series

thogard@xxxxxxxxxxxx (Tim Hogard) writes:

If it is possible to hide services from "svccfg archive" then that's
a bug which needs to be fixed.

I will build another example but right now I'm working on some other
issues. Such as... why can I kill -9 pkgadd inside a zone and
then pkgadd in the global zone won't work becaus it claims a pkgadd
is running in a zone? Even halting the zone doesn't fix it
and it has to be removed from the index file to allow pkgadd to
run again.

That's certainly a bad thing. But nothing the global administrator
cannot rectify (and "pkgadd -G" will skip the zone check)
The packaging tools require quite a bit of revamping and certainly
should be made independent of whatever tricks a zone administrator
pulls.

I've also been looking at the complexity of the core system. It
appears that the number of critical sections of critical files has
increased by about a factor of 100 from S9 so for the same application
reliability, so I need 100x better MTBF on the hardware to keep the
same MTBF on my application. Thats based purely on a low level
error and most hardware like disk drives don't commoly show single
bit errors, they die a screaming death and quickly take about 5%
of the bits out all at once.


100 fold increase appears to be taking some liberty with numbers.

Much of the rationale for SMF is increasing reliability; unfortunately,
this requires more software.

But in S10 services which die don't just silently go away as they
did in Solaris 9 and as they do on many other Unixes. They are
restarted. And Solaris 10 SMF allows a specification of all appropriate
service dependencies, so that a clean restart is possible; SMF can express
dependencies such as:

- when service X dies, restart Y before restarting X
- when service X dies, restart A, B, C after restarting X
- if process P from service X dies, kill all processes in service X
then restart X

Some of these required new Solaris mechanism (process contracts) and
there's really no way that "init.d on steriods" could have captured all
that.

Casper
.

Relevant Pages

  • Re: DNS 4010
    ... > services and restart a server. ... this resource record is corrupt or contains an invalid DNS name" This would ... Here is something you can try to see if it is the Netlogon ... for DNS only, then on the DC that both are using for DNS, change the zone to ...
    (microsoft.public.windows.server.dns)
  • Re: Cant use C: drive error checking tools
    ... I tried to prevent it from loading on startup ... FILE FOR DIRECT ACCESS on the screen after restart. ... >You are appreciating that you should set up chkdsk to ... Do you have Zone Alarm installed? ...
    (microsoft.public.windowsxp.perform_maintain)
  • Re: DNS 4010
    ... >> services and restart a server. ... > this resource record is corrupt or contains an invalid DNS name" This would ... Here is something you can try to see if it is the Netlogon ... > for DNS only, then on the DC that both are using for DNS, change the zone to ...
    (microsoft.public.windows.server.dns)
  • Re: 4010 Error continue
    ... Reed Wiedower posted their thoughts, ... > restart either of my two DNS servers the logs are filled with 4010 ... > Directory is functioning properly and reload the zone. ... I would first delete the netlogon.dns file, and restart the ...
    (microsoft.public.win2000.dns)
  • Unknown problem and/or symptoms.
    ... starts acting like it's on stand-by mode. ... I mean, it dies on me, the ... monitor turns off but the coolers are still working and i have to ... restart. ...
    (microsoft.public.windowsxp.help_and_support)