mandevil · Product Manager · ✭✭✭✭✭

deannich, Custom metrics have to be a single numeric return for DPA to handle it correctly. In this case, it seems likely that you'd have to create a metric for each database involved. If this metric doesn't fit into one of the existing categories, you can create your own custom category via the custom metric wizard. Once…

@xtraspecialj You may be hitting a known issue in the DPA integration with Orion. Do you mind submitting a support case? I can work with you via that channel on a buddy drop for what I believe is the same issue.

There should be a built in alert for this. If you go to alerts -> manage alerts, under "Wait Time" select the dropdown and select "Average Wait time for a single SQL". See if that is what you were thinking of.

First of all, I like that you've found some flexibility for the resource metrics. 8 ) Secondly, apart from some high level capacity planning trends, I would not recommend putting a lot of focus on resource metrics within that tab. I like to think of that view as my available library and a way to get accustomed to what I…

Great question and since DPA is cross DB platform, that resource metric makes more sense on the Oracle side. What I usually recommend for customers is to go to the resource view for the instance in question, click on the settings link under OS Memory Utilization, click on the "custom" radio button and adjust the…

What database engine are you looking at? SQL/Oracle/etc.? If SQL, then when you are in an instance, you can filter a couple of ways: Users running SQL against a specific database: * Click on Databases tab, then click on a day. In the upper left, you will see an interval drop down - choose 1 day. Click on the single…

Let me kick off the replies with a few thoughts: * We recommend getting DPA as close to the databases being monitoring as possible (and by close, I mean with least network latency possible) * We will be sending bursts of data across the network (primarily during our stats poll which occurs every 10 minutes) - so depending…

Here is a custom monitor for failed jobs: Job Failure You could turn off the OOTB alert and use this custom alert, modifying it for your conditional logic.

You might try something simple like: select (size/maxsize)*100 from sys.sysfiles where fileid = (select file_id from sys.dm_io_virtual_file_stats where database_name = 'tempdb') Make that into a custom alert with a single value and set thresholds at 80 and perhaps 90 percent?

You are spot on - there are limitations with older versions of SQL Server and XEs. Likely the best bet for that version is to set up a custom metric and alert for deadlock conditions (note, you need the metric for this alert). Something like: Custom Alert - Deadlocks

How many instances are you monitoring? Are they all on the downward trend? If yes, that's a good thing as these are waits being measured. Also, how long have you been monitoring? It will take a couple days to complete a baseline on which to base the trend on.

To my knowledge, you would have to click on the alert link taking you to the deadlock view in DPA which then would have details involved. The other option is to create a custom alert returning the data you are looking for. An example of setting up a custom alert can be found here.

How is your storage configured/architected? Are you hitting storage on a SAN? Local disk? As far as the objects spinning up every month, you may be referring to our monitored instances tables. We create a new set of tables for every monitored instance. That's what the _id represents and helps us map back to the COND table.…

David, what version of DPA are you on? With more recent versions (I believe 10.2 and up), on the blocking tab within 1 hour time frame, you can click on show last activity for the idle blocker session which may indeed catch the session and SQL causing the blocking. Keep in mind though that if the SQL ran sub-second, we may…

You might want to try adjusting the quickpoll_where_clause per the article link: Success Center You can limit the quickpoll to just grab wait data for one database. Keep permissions the same though as DPA will need access to DMVs still. All charts and other polls are driven by the QP, so limiting at that granular of a…

Are you monitoring SQL Server? Is the version 2012 or later? Here's a post discussing some things to try: DPA deadlock collection not enabled by default - SolarWinds Worldwide, LLC. Help and Support

Hmmm - that appears to be a new one for me... Since it may require back end data manipulation (deletes), I'd recommend opening a support ticket just in case we need engineering help.

Here's a great article showing how to exclude things from collection: SolarWinds Knowledge Base :: Excluding a SQL Statement from DPA Data Collection Note: please use with caution as sometimes things can start impacting you from a performance perspective that now you won't see because of the exclusion...

Mark, Idle blocking occurs when you have a session that opened a transaction (establishing a lock on a resource) and then did not commit or rollback explicitly. The transaction stays open, even though no work is currently being done. Imagine you do a begin transaction and issue an update to a table. The results will come…

Here is the query I see running in my environment. However, I think it depends on how the alert is configured as to how the ObjList is built out. A quick trace may reveal more details. DECLARE ObjList dbms_stats.ObjectTab; BEGIN dbms_stats.gather_database_stats(objlist=>ObjList, options=>'LIST AUTO'); FOR i in…

Currently, DPA can send alerts via SMTP and SNMP traps. As long as the ticketing system could consume one of those methods, you should be OK. We ship DPA with a MIB file that you can look at for the SNMP traps (it's in <install_dir>/iwc and called CONFIO-MIB.mib

Paul, you can just go into options -> under VMware column, either unregister (last ditch effort in my opinion) or update connection info (the update will allow you to change the PW to the correct one). If you unregister, you will lose all historical data for the Virtualization view. You can always register again which will…

Have you tried a multiple numeric alert? Not familiar with DB2 views, do you know where to get the database from? Can you join to another view for that info from mon_lockwaits? Something like: select c.tbsp_name, a.count(*) from sysibmadm.mon_lockwaits a, sysibmadm.mon_get_table b, sysibmadm.container_utilization c where…

Couple of thoughts here: * Looks like overall waits went down significantly, which is good (half what it was!) * Less time spent waiting for a transaction to commit, the more time for actually processing the next insert, update, or delete * Healthy systems should be spending their time in memory/CPU (anything else is…

It will change the top SQL. By excluding certain wait types, it will exclude SQL statements gathered in our quickpoll, so essentially, we won't "see" them any more.

Likely what's happening is that it fell off the list of the top 15 for the 30 day trend chart. All SQL statements listed are for the timeframe being viewed. It looks like that specific SQL has been supplanted by another one that's experiencing more waits... at least at the 30 day view.

On the primary only. Even though we are getting to the SQL Server instance via the AG listener, we're still monitoring at the overall instance. The AG listener will follow the primary replica, so we'll follow just like other apps.

At first glance, this would appear to be a contradiction. However, upon investigation, I found this article that I think explains in much detail what you are seeing. Great SQL Server Debates: Buffer Cache Hit Ratio - Simple Talk Spoiler alert, I think the read-ahead feature is saving this particular counter and may…

Just a heads up - engineering is taking a look at this. My first though is that everything before the statement itself is being thrown away to present as much of the actual query as possible since we know it will be truncated. As you state though, this could be a disservice in a case like you have here...

There is a pretty slick way to get this, but it is an option that should be used with support guidance. Please open a ticket with them and they'll be happy to help out!