Monthly Maintenance (Reboot) Needed?

[Courtney Bush]

Last year, we migrated (@75) interfaces from a Market A running 3.8.1 to a new corporate 5.3 engine running on AIX 5.2.  Since then, we have implemented HA.  This year, we will be migrating another market’s interfaces, Market B, doubling our current number of interfaces on the corporate engine.  Market A is used to performing a monthly reboot, which creates a downtime of over an hour.  Now that we’re adding Market B (who is not accustomed to a monthly maintenance reboot), we want to reassess if this monthly reboot is even needed – given the engine version, HA, stability of the OS, etc.  Given our confidence in the corporate engine, it will be difficult to coordinate a multi-market downtime and it is hard to justify to the Corporate CIO that with all of the money invested with HA, that a 1+ hour downtime is needed for both markets on a monthly basis.

My question:  Do you perform a regularly scheduled maintenance period?  

If so, what does it gain you and how often is this scheduled for?

If not, what metrics do you suggest we monitor that would indicate to us that one was needed?

Thank you for your time in responding.

Topic

[James Cobane]

Courtney,

Since you are running on AIX there really is no need for a monthly (or regularly scheduled) re-boot of the interface engine.  In fact, in past postings to this forum, there were always “bragging rights” to whoever had the longest time between re-boots.  The only time we re-boot is when it is required for installing maintenance levels or upgrades; we are running Cloverleaf 5.4.1 on AIX 5.3.  Tell Market A that they can enjoy the extra uptime.

Jim Cobane

Henry Ford Health

[Grady Stephens]

I know when we were on 3.8 if we did not re-boot periodically, we had problems.  We have since moved to 5.3 and continue to re-boot but I would like to know if the re-boot is necessary now also.

[Kevin Scantlan]

I would say if you have to reboot then chances are that you have some memory leaks.  We never reboot except to OS maintenance.

[Russ Ross]

I just manually invoked a HACMP fail-over this past Sunday at 2 AM and did a reboot for the purpose of installing the updates for daylight savings time on our various LPAR’s.

Our Cloverleaf prodcution LPAR had been up for 381 days running AIX 5.2 with HACMP and Cloverleaf 5.2.1P2.

I think that is the longest stretch we have ever been up without any down time, yet the phones rang and people cried even at 2 AM on Sunday and I had to remind them they had been spoiled and had forgotten I do have the right to down time at 2 AM on Sunday(s) when needed.

We have a large number of interfaces and sites and HA fail-over took an hour when our box was not powerfull enough for us to feel comfortable starting sites up in parallel.

Now with our current powerful hardware we start sites up in staggered parallel and it takes about 15 minutes to fail-over.

It almost seems we could stay up indefinitely but like James Cobane pointed out, we never get the chance to find out because some upgrade to the OS, HA, SNA or Cloverleaf will occur.

When I’m not too lazy, I like to fail-over and reboot after 6 months because I don’t want potential problems to go unoticed for too long and our backups start to become stale after 6 months.

Russ Ross
RussRoss318@gmail.com

[Author]

Replies