I'd like to revisit the age old discussion about the best way to do drive
cleaning. I'm working at a site with a large number of tape libraries of
various sizes, mainly from StorageTek, but also some older ADIC units. The
drives are DLT7000 gradually migrating to LTO. They currently use NetWorker
cleaning on most of these and we are trying to establish the best strategy.
Here is my attempt to be objective about the pros and cons of the two
possible approaches, as well as my own opinion. I would be very interested
to hear what the NetWorker community think about this emotive issue. Some
views from Legato staff would also be welcome.

NetWorker controlled cleaning.
NetWorker is configured to clean drives at set intervals of elapsed time.
The cleaning tape(s) is/are stored in a designated tape slot within the
range normally used for data tapes. The number of cleaning tape uses is
configured within NetWorker. When a cleaning tape is used up, a
notification can be sent from NetWorker.

Library firmware controlled cleaning.
The library firmware is configured to clean drives on demand. The DLT or
LTO drives are sophisticated enough to send a signal to the library when
cleaning is needed, and the library automatically loads and unloads the
cleaning tape at exactly this time. The cleaning tape(s) is/are stored in
dedicated cleaning slots (library dependent). The number of cleaning tape
uses is configured in the library firmware. When a cleaning tape is used
up, most libraries withdraw the tape to the CAP and display a message on
the library front panel, this may be accessible remotely if the library has
remote management software and a network connection.

What is wrong with NetWorker controlled cleaning?
* The cleaning is based on an elapsed time since the last clean, e.g. two
weeks. This takes no account of the drive usage in this time, a heavily
used storage node drive may have had perhaps 168 hours of use in that time,
whereas a drive attached to an application server may have only had perhaps
14 hours use. It is highly likely that the former case would have required
a clean before that time, but unlikely that the latter case is anywhere
near to needing a clean.

* Cleaning a drive before it is required is wasteful of cleaning tapes and
can be bad for the drive. DLT drives need infrequent cleaning, LTO drives
rarely need cleaning at all.

* A heavily used drive may require cleaning before the cleaning interval is
reached. It will not get cleaned on demand, and may cause errors or damage
to data tapes. You can clean manually if there are errors but this is not
ideal on a remote site because unless you can actually see the library you
don't know whether the error was caused by the need to clean or some other
problem.

* One or more data slots are lost due to the need to store cleaning tapes.

* Because cleaning tapes are in data slots, you must take care to specify a
slot range to avoid the cleaning tapes when inventorying the jukebox.

What is wrong with library controlled cleaning?
* NetWorker may show an error on loading a tape when the library is
cleaning a drive. Usually these errors are recoverable without user
intervention.

* Cleaning tape usage data is only available from the library front panel.
Some libraries have remote monitoring software available, but this requires
a network connection which may not have been set up. Even with remote
monitoring software, it requires pro-activity, to go and check usage rather
than being automatically alerted.

* When a drive goes bad, it may repeatedly request cleaning and use up a
cleaning tape. NetWorker cleaning would just show errors. Which is worse?

* Not all libraries support this feature.

Legato Support always recommended using library firmware cleaning where
available in pre 6.x days. Now it seems that they are recommending
NetWorker cleaning instead. I don't understand the reason for this U-turn,
since all the disadvantages listed above still apply. I firmly believe that
library firmware cleaning is superior to NetWorker cleaning.

--
Note: To sign off this list, send a "signoff networker" command via email
to listserv AT listmail.temple DOT edu or visit the list's Web site at
http://listmail.temple.edu/archives/networker.html where you can
also view and post messages to the list.
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=