Skip to Content.
Sympa Menu

perfsonar-dev - Request for less-cryptic error messages

Subject: perfsonar development work

List archive

Request for less-cryptic error messages


Chronological Thread 
  • From: Jason Zurawski <>
  • To: "" <>
  • Subject: Request for less-cryptic error messages
  • Date: Wed, 08 Jul 2009 17:25:53 -0400
  • Organization: Internet2

All;

This is a request for the MDM LS specifically, but the issue pertains to all services that pass along eventTypes and datum-messages regarding system errors.

My dilemma: Trying to do what I thought was a simple query to both MDM and pS instances of the hLS for a web page. The query I sent:


/nmwg:store[@type="LSStore"]/nmwg:metadata

The pS hLS responded ok, but the MDM hLS gave me this initially:

eT: error.common.storage.xmldb.open</nmwg:eventType>
datum: XQuery by HTTP failed. Could not connect to eXist via pure http, nested exception was: class org.perfsonar.service.base.storage.xmldb.exist.rest.ExistDbErrorException: eXist DB responsed with HTTP error code code [202] additional message was [Accepted]</nmwgr:datum>

I would have pegged this as a configuration error with the service since the eventType appears to be 'system' related. I figured out quickly that every MDM LS out in the wild responded like this, so it appears to be my fault. After some poking I realized I should have sent this variant instead:

declare namespace nmwg="http://ggf.org/ns/nmwg/base/2.0/";;

/nmwg:store[@type="LSStore"]/nmwg:metadata

My thoughts:

1) I am sure the above message comes from eXist directly, which is unfortunately vauge. Since my problem was really a syntax/formulation error with the XQuery/XPath statement, I would much rather see this sort of error message (cleaned up a little of course): "Hey Dummy, learn XPath!".

Perhaps we need to spend some time decoding the error messages that would come from supporting software (e.g. RRD, XMLDB, eXist, MySQL) into more user/protocol friendly descriptions and having the services simply look these up. Standardizing across services/implementations would be a bonus.

2) Even though this wasn't a system error, would a system error be returned via the message protocol? I think we touched on this years ago but where do we draw the line between what is good to tell the users and what is good to tell the sysadmin via logging.

Thanks;

-jason

P.S. New *development* version of the 'tree' GUI that has a feature to dump hLS contents:

http://dc211.internet2.edu/cgi-bin/pA/tree.cgi



Archive powered by MHonArc 2.6.16.

Top of Page