Distributed Lucene with Persistence Counters
NCache provides various cache server counters to monitor the performance of cache servers. You can monitor all these counters for a specific cache on any Windows machine using either NCache Web Monitor or PerfMon tool.
These counters include:
Counters | Description |
---|---|
Indexes |
The total number of indexes created by user, each identified by a unique name. An index is created when an NCacheDirectory is opened, if an index with same name is already present, then a new one won't be created. |
Index Size |
The size of all indexes on that particular node. Main and replica nodes are differentiated, so in case of POR topology, main node will only consider the 'main' portions of all indexes and replica will follow suit. |
Documents |
Total number of documents on that node. Ideally, the number of documents on replica node should be same as its corresponding main node. |
Commits |
Total number of commits that have occurred on that node. Commits are not persisted, so after a node restart they are reset, also in case of state transfer, commit count is not carried over. Commits include all commits initiated by user, as well as those done automatically by flush policy. |
Additions/sec |
Number of document insertions per second for a node. |
Updates/sec |
Number of update operations on documents per second for a node. |
Deletes/sec |
Number of all types of delete operations per second for a node. Since delete operations are query based and there are multiple writers performing the delete operation, it is difficult to calculate the total number of deleted documents, so one delete operation might be deleting just 1 document or a 1000. |
Fetches/sec |
Number of 'get' type operations performed per second on a node. For example, GetDocCount() is a fetch operation. |
Searches/sec |
Number of query based operations performed per second on a node. Strictly speaking, not all operations performed on IndexSearcher are considered search operations, e.g. CollectionStatistics(), which returns stats for a particular field is a fetch operation. |
Requests/sec |
Total number of all types of requests received by the Lucene module. |
State Transfer/sec |
The number of bytes being transferred per second upon state transfer. |
Avg ms/add |
The average time in millisecond an add operation takes. |
Avg ms/update |
The average time in millisecond an update operation takes. |
Avg ms/delete |
The average time in millisecond a delete operation takes. |
Avg ms/fetch |
The average time in millisecond a fetch operation takes. |
Avg ms/search |
The average time in millisecond a search operation takes. |
Buckets to Send |
The number of buckets a node has to send (if any) to other nodes. A scenario to see this in action, is to create a two node cluster (POR) and just start the first node. You should see 50 buckets which started node has to send to the 2nd (stopped) node, upon starting 2nd node, this will go down to zero. |
Buckets to Receive |
The number of buckets a node has to receive (if any) from other nodes. Since state transfer occurs very quickly, it is difficult to see this updating, as the node will receive all of its buckets as soon as it is started. But the above scenario will suffice, just keep an eye on the 2nd node. |
Replication Queue Count |
The number of 'operations' the main node has to replicate to its replica node, this will include all types of operations. |
Live Instances |
The total number of live instances (objects such as Writers, Readers and Searchers) currently instantiated by the client. Disposed objects are removed (objects are disposed on client disconnect as well). Unlike commits, live instances are carried over to the new nodes. |