the command is under install_directory/bin, actualy , it is base type of operation system.

Since no one seems to know let me document what I have discovered.

Firstly there is a mistake in the man pages, where it says -g/-G program name replace that with -o/-O program name.  There is another switch, -z/-Z which seems to convert hostnames into an internal name (they come out host#).

Secondly the format of the data you get out is like this for the default output:

77561 12/18/2012 06:01:31 PM  2 0 0 4146657120 25773 0 hostname savegrp command 'savefs command'' for client xxxyyy exited with return code 1

Fields in the above are:

message ID, timestamp, severity, category, error no, thread id, process id, activity id, hostname, program name, message

The basic functionality of the switches is that each field is represented by a letter and:

-(lower case letter) = exclude field in output

-(upper case letter) value = include lines where the specified field matches the specified value

The fields mean the following:

m message ID = code mapping to the logged message (the last field), things like waiting for job to complete etc

     timestamp = time error logged

s   severity = this returns fields EQUAL OR MORE to the specified value, the higher the level the more severe the problems returned (and less info returned)(0-5 for notification, information, alert, warning, critical, emergency)

c    category = there appears to be two valid values, 0 is regular server logging, 3 is alerts relating to permissions

e    error = note here the -e switch excludes the error code, but the -E switch specifies the end time; it appears you cannot specify an error code in query

t      thread ID = specific thread ID

process ID = process ID

a    activity ID = seems to always be 0

h    hostname = this seems to always be the hostname the log file is on

p    program name = savegrp, nsrmmdbd, nsrjobsd etc

       message = the message corresponding to the message ID which may have specific things relating the message such as client name, saveset name and so on.

man nsr_render_log will give you enought... as to what is the meaning of what you get, it will take some digging, but some are message ids related to which daemon is generating them and rest is pretty much easy to figure out.  One of the most disturbing things I find about nsr_render_log is fact that if I use -G it does not capture completion as that is given by another owner which is not covered by -G grep. 

Other useful things to know:

You can query on a particular client with -J

You can query specific dates using nsr_get_date format, one useful way of doing this is:

nsr_render_log -S "yesterday 09:00:00" -E "today 09:00:00" /nsr/logs/daemon.raw

- will return all backups from 9AM yesterday morning to 9AM this morning.

You can set a character to use as a separator:

-x 'c \t'    (will tab separate the fields)

A list of message IDs would be useful if I can find that!

Yes, -S and -E are the only switches I use with nsr_render_log... others I didn't find useful to use and if I need to get rid of them I simply post process it.

Hello, you may be happy to hear that this is under review at the moment; many of the fields you mentioned are currently not used to any great effect (severity, category, error, activity).

Also, please note that while the intent as been to ensure a unique message ID for all messages, this is currently not a guarantee either.

man nsr_render_log, or the Command Reference should provide you all you need to know; meanwhile, we will be working hard to expand upon the potential of the format for better usability.

HTH, James.

Command Reference for nsr_render log: http://powerlink.emc.com/km/live1/en_US/Offering_Technical/Technical_Documentation/300-999-721.pdf#page=373

Error Message Guide: http://powerlink.emc.com/km/live1/en_US/Offering_Technical/Technical_Documentation/300-999-724.pdf (includes MsgIds - but as previous, not truly unique)