IMHO, I think the drivers are fine. It was a dumb mistake of mine to use sessions as connections and not as connection pools. What was harder to figure, in my opinion, was that too many connections from the client would increase the amount of file descriptors used by the server. I didn't know Linux open a FD for each connection received and honestly I still don't know much about the details of this. When I got a "too many open files" error it took a good while to think about checking the connections. I think the documentation could point this fact, it would help other people with the same problem. There could be something talking about it here: http://www.datastax.com/documentation/cassandra/2.0/cassandra/troubleshooting/trblshootTooManyFiles_r.html
[]s 2014-08-09 12:55 GMT-03:00 Jack Krupansky <j...@basetechnology.com>: > Maybe the drivers should have two modes: few sessions, and lots of > sessions. The former would give you a developer-friendly driver error if > you leave more than say a dozen or two dozen sessions open (or whatever is > considered a best practice for parallel threads in a client), on the theory > that you probably used the anti-pattern of failing to reuse sessions. The > latter would be more for expert apps that have some good reason for having > hundreds or thousands of simultaneous sessions open. Whether the latter > also has some (configurable) limit that is simply a lot higher than the > former or is unlimited, is probably not so important. Or, maybe, simply > have a single limit, without the modes and default it to 10 or 25 or some > other relatively low number for “normal” apps. > > This would be more developer-friendly, for both new and “normal” > developers... I think. > > -- Jack Krupansky > > *From:* Marcelo Elias Del Valle <marc...@s1mbi0se.com.br> > *Sent:* Saturday, August 9, 2014 12:41 AM > *To:* user@cassandra.apache.org > *Subject:* Re: too many open files > > Indeed, that was my mistake, that was exactly what we were doing in the > code. > []s > > > 2014-08-09 0:56 GMT-03:00 Brian Zhang <yikebo...@gmail.com>: > >> For cassandra driver,session is just like database connection pool,it >> maybe contains many tcp connections,if you create a new session every >> time,more and more tcp connections will be created,till surpass the max >> file description limit of os. >> >> You should create one session,use it repeatedly ,session can manage >> connections automatically,create new connection or close old connection for >> your requests. >> >> 在 2014年8月9日，6:52，Redmumba <redmu...@gmail.com> 写道： >> >> Just to chime in, I also ran into this issue when I was migrating to >> the Datastax client. Instead of reusing the session, I was opening a new >> session each time. For some reason, even though I was still closing the >> session on the client side, I was getting the same error. >> >> Plus, the only way I could recover was by restarting Cassandra. I did not >> really see the connections timeout over a period of a few minutes. >> >> Andrew >> On Aug 8, 2014 3:19 PM, "Andrey Ilinykh" <ailin...@gmail.com> wrote: >> >>> You may have this problem if your client doesn't reuse the connection >>> but opens new every type. So, run netstat and check the number of >>> established connections. This number should not be big. >>> >>> Thank you, >>> Andrey >>> >>> >>> On Fri, Aug 8, 2014 at 12:35 PM, Marcelo Elias Del Valle < >>> marc...@s1mbi0se.com.br> wrote: >>> >>>> Hi, >>>> >>>> I am using Cassandra 2.0.9 running on Debian Wheezy, and I am having >>>> "too many open files" exceptions when I try to perform a large number of >>>> operations in my 10 node cluster. >>>> >>>> I saw the documentation >>>> http://www.datastax.com/documentation/cassandra/2.0/cassandra/troubleshooting/trblshootTooManyFiles_r.html >>>> and I have set everything to the recommended settings, but I keep getting >>>> the errors. >>>> >>>> In the documentation it says: "Another, much less likely possibility, >>>> is a file descriptor leak in Cassandra. Run lsof -n | grep java to >>>> check that the number of file descriptors opened by Java is reasonable and >>>> reports the error if the number is greater than a few thousand." >>>> >>>> I guess it's not the case, or else a lot of people would be complaining >>>> about it, but I am not sure what I could do to solve the problem. >>>> >>>> Any hint about how to solve it? >>>> >>>> My client is written in python and uses Cassandra Python Driver. Here >>>> are the exceptions I am having in the client: >>>> [s1log] 2014-08-08 12:16:09,631 - cassandra.pool - WARNING - Error >>>> attempting to reconnect to 200.200.200.151, scheduling retry in 600.0 >>>> seconds: [Errno 24] Too many open files >>>> [s1log] 2014-08-08 12:16:09,632 - cassandra.pool - WARNING - Error >>>> attempting to reconnect to 200.200.200.142, scheduling retry in 600.0 >>>> seconds: [Errno 24] Too many open files >>>> [s1log] 2014-08-08 12:16:09,633 - cassandra.pool - WARNING - Error >>>> attempting to reconnect to 200.200.200.143, scheduling retry in 600.0 >>>> seconds: [Errno 24] Too many open files >>>> [s1log] 2014-08-08 12:16:09,634 - cassandra.pool - WARNING - Error >>>> attempting to reconnect to 200.200.200.142, scheduling retry in 600.0 >>>> seconds: [Errno 24] Too many open files >>>> [s1log] 2014-08-08 12:16:09,634 - cassandra.pool - WARNING - Error >>>> attempting to reconnect to 200.200.200.145, scheduling retry in 600.0 >>>> seconds: [Errno 24] Too many open files >>>> [s1log] 2014-08-08 12:16:09,635 - cassandra.pool - WARNING - Error >>>> attempting to reconnect to 200.200.200.144, scheduling retry in 600.0 >>>> seconds: [Errno 24] Too many open files >>>> [s1log] 2014-08-08 12:16:09,635 - cassandra.pool - WARNING - Error >>>> attempting to reconnect to 200.200.200.148, scheduling retry in 600.0 >>>> seconds: [Errno 24] Too many open files >>>> [s1log] 2014-08-08 12:16:09,732 - cassandra.pool - WARNING - Error >>>> attempting to reconnect to 200.200.200.146, scheduling retry in 600.0 >>>> seconds: [Errno 24] Too many open files >>>> [s1log] 2014-08-08 12:16:09,733 - cassandra.pool - WARNING - Error >>>> attempting to reconnect to 200.200.200.77, scheduling retry in 600.0 >>>> seconds: [Errno 24] Too many open files >>>> [s1log] 2014-08-08 12:16:09,734 - cassandra.pool - WARNING - Error >>>> attempting to reconnect to 200.200.200.76, scheduling retry in 600.0 >>>> seconds: [Errno 24] Too many open files >>>> [s1log] 2014-08-08 12:16:09,734 - cassandra.pool - WARNING - Error >>>> attempting to reconnect to 200.200.200.75, scheduling retry in 600.0 >>>> seconds: [Errno 24] Too many open files >>>> [s1log] 2014-08-08 12:16:09,735 - cassandra.pool - WARNING - Error >>>> attempting to reconnect to 200.200.200.142, scheduling retry in 600.0 >>>> seconds: [Errno 24] Too many open files >>>> [s1log] 2014-08-08 12:16:09,736 - cassandra.pool - WARNING - Error >>>> attempting to reconnect to 200.200.200.185, scheduling retry in 600.0 >>>> seconds: [Errno 24] Too many open files >>>> [s1log] 2014-08-08 12:16:09,942 - cassandra.pool - WARNING - Error >>>> attempting to reconnect to 200.200.200.144, scheduling retry in 512.0 >>>> seconds: Timed out connecting to 200.200.200.144 >>>> [s1log] 2014-08-08 12:16:09,998 - cassandra.pool - WARNING - Error >>>> attempting to reconnect to 200.200.200.77, scheduling retry in 512.0 >>>> seconds: Timed out connecting to 200.200.200.77 >>>> >>>> >>>> And here is the exception I am having in the server: >>>> >>>> WARN [Native-Transport-Requests:163] 2014-08-08 14:27:30,499 >>>> BatchStatement.java (line 223) Batch of prepared statements for >>>> [identification.entity_lookup, identification.entity] is of size 25216, >>>> exceeding specified threshold of 5120 by 20096. >>>> ERROR [Native-Transport-Requests:150] 2014-08-08 14:27:31,611 >>>> ErrorMessage.java (line 222) Unexpected exception during request >>>> java.io.IOException: Connection reset by peer >>>> at sun.nio.ch.FileDispatcherImpl.read0(Native Method) >>>> at sun.nio.ch.SocketDispatcher.read(SocketDispatcher.java:39) >>>> at sun.nio.ch.IOUtil.readIntoNativeBuffer(IOUtil.java:223) >>>> at sun.nio.ch.IOUtil.read(IOUtil.java:192) >>>> at sun.nio.ch.SocketChannelImpl.read(SocketChannelImpl.java:375) >>>> at >>>> org.jboss.netty.channel.socket.nio.NioWorker.read(NioWorker.java:64) >>>> at >>>> org.jboss.netty.channel.socket.nio.AbstractNioWorker.process(AbstractNioWorker.java:109) >>>> at >>>> org.jboss.netty.channel.socket.nio.AbstractNioSelector.run(AbstractNioSelector.java:312) >>>> at >>>> org.jboss.netty.channel.socket.nio.AbstractNioWorker.run(AbstractNioWorker.java:90) >>>> at >>>> org.jboss.netty.channel.socket.nio.NioWorker.run(NioWorker.java:178) >>>> at >>>> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142) >>>> at >>>> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617) >>>> at java.lang.Thread.run(Thread.java:745) >>>> >>>> Here is the amount of open files before and after I restart Cassandra: >>>> >>>> root@h:/etc/security/limits.d# lsof -n | grep java | wc -l >>>> 936580 >>>> root@h:/etc/security/limits.d# sudo service cassandra restart >>>> [ ok ] Restarting Cassandra: cassandra. >>>> root@h:/etc/security/limits.d# lsof -n | grep java | wc -l >>>> 80295 >>>> >>>> >>>> Best regards, >>>> Marcelo Valle. >>>> >>> >>> >> >> > >
