11.7.1 introduces database diagnostics events.
The documentation shows an example setting
-diagEvent locktable:-15,bithold:4,syserror:8
are those recommended settings or just an example ?
I'd go for level 3 (detailed info + protrace) on all events to start with
-diagEvent locktable:3,bithold:3,syserror:3
It doesn't seem like a feature where someone else can provide all the answers for you. The event levels determine how much information you will get and how often you will get it. Only you can decide what is enough for you, or too much.
In answer to your question, I would say that the example is just an example, not a recommendation. Event levels 4-7 (and 12-15) involve running an executable when the event fires. If you don't have such a program to run then those levels aren't useful to you. If you want to receive diagnostic information for every event and not just when the event is fatal to the database then the negative numbers aren't appropriate for you. If you don't want to receive information about all events when only one event fires then you don't need levels 8-15.
I haven't tried to use this feature but based on what I know so far I would suggest starting with event level 3, assuming you have self-service clients and you're not on a Windows back end. Protraces (level 3 and 11) are Unix-only. You won't get protraces for remote clients; if your clients are remote, use 2. Once you collect some data you can decide whether the extra detail information is useful to you for diagnosing the problem. If it isn't then you can reduce the level for that event to 1 (summary only).
In short, experiment with it, incorporate your preferences, and see what works for you. I'd be interested to hear your feedback once you start to see some data.
It doesn't seem like a feature where someone else can provide all the answers for you. The event levels determine how much information you will get and how often you will get it. Only you can decide what is enough for you, or too much.
In answer to your question, I would say that the example is just an example, not a recommendation. Event levels 4-7 (and 12-15) involve running an executable when the event fires. If you don't have such a program to run then those levels aren't useful to you. If you want to receive diagnostic information for every event and not just when the event is fatal to the database then the negative numbers aren't appropriate for you. If you don't want to receive information about all events when only one event fires then you don't need levels 8-15.
I haven't tried to use this feature but based on what I know so far I would suggest starting with event level 3, assuming you have self-service clients and you're not on a Windows back end. Protraces (level 3 and 11) are Unix-only. You won't get protraces for remote clients; if your clients are remote, use 2. Once you collect some data you can decide whether the extra detail information is useful to you for diagnosing the problem. If it isn't then you can reduce the level for that event to 1 (summary only).
In short, experiment with it, incorporate your preferences, and see what works for you. I'd be interested to hear your feedback once you start to see some data.
Can these parameters be activated online?
Hi Rob,
Thank you, your answer confirms my idea of starting with level 3.
Yes, you can update the values using promon or the _dbparams VST.
Check rdbms_diagnostic_data.pdf (from OE1171SP_SupportingDocs.zip)
knowledgebase.progress.com/.../OpenEdge-Service-Pack-11-7-1-Is-Now-Available
or Richard Banville's "Database Diagnostics" presentation:
Thanks George!