Hi! > Hello Bobby,
Hey! > Does the ONMS architecture take my multiple rows and sum them, merge > them, etc. before saving them to the RRD files? Or, does ONMS only > persist the first entry in the CollectionSet? > > ONMS uses an implementation of the CollectionSetVisitor interface > called BasePersister to persist the data on RRD/JRB files. > The CollectionSet will be persisted as it is, so its content must be > correct, otherwise some weirdness may happen on the filesystem > structure for the RRD files and the RRD itself. I'm not sure this answers my question. Or, if its all just in the Java Code, I'll have to go re-read. My XMP collector can be thought of in similar terms to SNMP. XMP is a SSL/TCP protocol that uses the a modified SNMP SMI and similar MIBs. XMP has a richer table model along with several other enhancements plus SSL for authentication and privacy. I model the system's process table as a MIB table. Lets say I collect stats for processes that match "postgres". My XMP Collector issues an XMP SelectTableRequest with key="postgres" and my agent will send back all the processes matching it. For my ONMS system, it returns back the set of processes (rows in process table) matching this. My collector then adds each row to the CollectionSet that is then returned. How is a this data then persisted to a JRB file? Thats my question. Is the data summed? Or, is each row just written on top of the proceeding row and what is finally stored is the last row to be read? > Some time ago, I've modified the JDBCCollector and the WMICollector to > be sure that only one instance is created for the "node-level" > resource, and reuse it when populating the CollectionSet. Without this > change, multiple node-level resources can be created which is > theoretically wrong, because a node can only have one node-level > resource. That's why it is called node-level ;) > Here is the summary of the problem I fixed: > [2]http://issues.opennms.org/browse/NMS-5219 I'm not sure I understand this based on the summary. Or, I don't think its relevant. My collector is already collecting scalar and tabular data and for filesystems and interfaces the data is being persisted correctly. > I didn't apply the same fix for the XMP Collector because I'm not > familiar with it, so I hope you can do it. > The rest of the metrics must be considered tabular data > (GenericIndexResource). These kind of resources are stored inside a > folder. This folder typically contains the name of the resource type. > Of course, inside each folder, there are one folder for each resource > that must be unique. Finally, inside each resource folder, you'll see > the JRB/RRD files for the metrics. The structure of the JRB/RRD depends > on the storeByGroup variable. > > Hope that helps. I essentially have almost no clue what you are talking about. I think my collector is coded correctly but I'm not sure what is happening behind the scenes. Process stats could be a slightly different beast. Interfaces and filesystems are correctly collected and persisted in the XmpCollector. I may need to add an application table to my MIB and do the summation work in my agent as an application is modeled as a collection of processes. Thanks, Bobby
pgpJY2vOrR4Ds.pgp
Description: PGP signature
------------------------------------------------------------------------------
_______________________________________________ Please read the OpenNMS Mailing List FAQ: http://www.opennms.org/index.php/Mailing_List_FAQ opennms-devel mailing list To *unsubscribe* or change your subscription options, see the bottom of this page: https://lists.sourceforge.net/lists/listinfo/opennms-devel
