I configure the connection pool before anything is inserted (initialization
is in a static constructor).
So something like:
static OPartitionedDatabasePool pool = new OPartitionedDatabasePool(URL,
USER, PASSWOORD);
Then I have a method like:
ODatabaseDocumentTx getConnection()
{
ODatabaseDocumentTx conn = pool.acquire();
// register any hooks...
return conn;
}
- So connection pool is initialized statically.
- My test API gets a connection in the constructor and holds that
connection privately.
- One insert operation is performed on the test API.
- The test API is closed, inside the close the database is closed.
So the database is acquired via the connection pool, one document is
inserted, and the connection is closed. This is repeated hundreds of times,
which eventually slows down...
On Monday, September 14, 2015 at 9:20:35 AM UTC-7, Enrico Risa wrote:
>
> Where you configure the OPartitionedDatabasePool?
>
> Do you call db.close() ?
>
> 2015-09-14 18:00 GMT+02:00 Bryan <[email protected] <javascript:>>:
>
>> I am using the embedded, plocal database.
>> I insert about 1000 documents before I see insertion time start to
>> increase, although I want to insert much more than that. I'm not passing
>> anything to JVM arguments so I'm using default OrientDB configuration for
>> memory. One thing I should note, I don't know what thread will call my API
>> so re-use/open a connection from the OPartitionedDatabasePool every time an
>> insert is performed (not sure if that matters).
>>
>> Essentially, I do this...
>>
>> public void writeLogEntry(ODocument log, String message, Date timestamp,
>> int entrySize)
>> {
>> // get connection from OPartitionedDatabasePool...
>> ODatabaseDocumentTx db = this.dataCore.getDbConnection();
>> db.begin();
>>
>> ODocument logEntryDocument = new ODocument("log_item");
>>
>> // Set the log event fields for the new log event.
>> logEntryDocument.field("timestamp", timestamp);
>> logEntryDocument.field("message", message);
>> logEntryDocument.save();
>>
>>
>> long currentLogSize = log.field("log_size");
>>
>> // Update the log size by adding the size of the log event.
>> log.field("log_size", (currentLogSize + entrySize));
>>
>> // Add the log entry to the LINKSET in the log
>> Set<OIdentifiable> logEntries = log.field("log_events");
>> logEntries.add(logEntryDocument.getIdentity());
>> log.save();
>>
>> db.commit();
>> }
>>
>>
>>
>> On Monday, September 14, 2015 at 8:42:09 AM UTC-7, Enrico Risa wrote:
>>>
>>> Hi Bryan
>>>
>>> Are you inserting through plocal?
>>>
>>> how many records do you have to insert?
>>>
>>> After how many records do you see the slow down?
>>>
>>> Did you check the memory of your inserting batch?
>>>
>>>
>>> Enrico
>>>
>>> 2015-09-14 17:37 GMT+02:00 Bryan <[email protected]>:
>>>
>>>> Okay, I created issue #4965
>>>> <https://github.com/orientechnologies/orientdb/issues/4965>.
>>>>
>>>> What about the 2nd issue I have? Is there some reason inserting things
>>>> keeps getting slower and slower until I restart OrientDB?
>>>>
>>>> On Friday, September 11, 2015 at 10:50:49 AM UTC-7, Bryan wrote:
>>>>>
>>>>> Hi all,
>>>>>
>>>>> I have been experimenting with an OrientDB setup (version 2.1.0 -
>>>>> plocal connection) and noticed a very large difference in performance
>>>>> when
>>>>> passing in an OClass parameter to a LINKSET vs leaving it out.
>>>>> Example:
>>>>>
>>>>> OClass logItem = schema.createClass("log_item");
>>>>> logItem.createProperty("timestamp", OType.DATETIME);
>>>>> logItem.createProperty("message", OType.STRING);
>>>>>
>>>>> OClass log = schema.createClass("log");
>>>>> log.createProperty("name", OType.STRING);
>>>>>
>>>>> and
>>>>>
>>>>> log.createProperty("log_items", OType.LINKSET, logItem);
>>>>> VS
>>>>> log.createProperty("log_items", OType.LINKSET);
>>>>>
>>>>> I tested inserting 1 log and 1000 log items into the log.
>>>>> i.e.
>>>>> // log = Get the log...
>>>>> ODocument logItem = new ODocument("log_item");
>>>>> logItem.field("timestamp", new Date());
>>>>> logItem.field("message", "Hello World");
>>>>> logItem.save();
>>>>>
>>>>>
>>>>> Set<OIdentifiable> logItems = log.field("log_items");
>>>>> logItems.add(logItem.getIdentity());
>>>>> log.save();
>>>>>
>>>>>
>>>>> WITH OClass param => 38.8 seconds
>>>>> WITHOUT OClass param => 1.5 seconds
>>>>>
>>>>> So really I have 2 questions:
>>>>> 1) Why are updates so much faster when I DON'T add the OClass
>>>>> parameter. Is there a reason why I want to leave it in?
>>>>>
>>>>> 2) I'm experiencing decrease in insertion speed as the number of
>>>>> documents grows. If I shutdown OrientDB and start again I see speed go up
>>>>> temporarily until I insert more records and it continues to slow down...
>>>>> I've tried committing after every ".save()". Am I missing something?
>>>>>
>>>>> Thanks!
>>>>>
>>>> --
>>>>
>>>> ---
>>>> You received this message because you are subscribed to the Google
>>>> Groups "OrientDB" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send
>>>> an email to [email protected].
>>>> For more options, visit https://groups.google.com/d/optout.
>>>>
>>>
>>> --
>>
>> ---
>> You received this message because you are subscribed to the Google Groups
>> "OrientDB" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to [email protected] <javascript:>.
>> For more options, visit https://groups.google.com/d/optout.
>>
>
>
--
---
You received this message because you are subscribed to the Google Groups
"OrientDB" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
