Aquiles Discussions Rss Feed

New Post: Socked Exception When Mutate

JamesYYang — Wed, 16 Jan 2013 04:42:31 GMT

We get a socket exception when mutate value:

[SocketException (0x274c): A connection attempt failed because the connected party did not properly respond after a period of time, or established connection failed because connected host has failed to respond]
   System.Net.Sockets.Socket.Receive(Byte[] buffer, Int32 offset, Int32 size, SocketFlags socketFlags) +6210712
   System.Net.Sockets.NetworkStream.Read(Byte[] buffer, Int32 offset, Int32 size) +134

[IOException: Unable to read data from the transport connection: A connection attempt failed because the connected party did not properly respond after a period of time, or established connection failed because connected host has failed to respond.]
   System.Net.Sockets.NetworkStream.Read(Byte[] buffer, Int32 offset, Int32 size) +300
   Thrift.Transport.TStreamTransport.Read(Byte[] buf, Int32 off, Int32 len) +141
   Thrift.Transport.TTransport.ReadAll(Byte[] buf, Int32 off, Int32 len) +90
   Thrift.Transport.TFramedTransport.ReadFrame() +88
   Thrift.Transport.TFramedTransport.Read(Byte[] buf, Int32 off, Int32 len) +143
   Thrift.Transport.TTransport.ReadAll(Byte[] buf, Int32 off, Int32 len) +90
   Thrift.Protocol.TBinaryProtocol.ReadAll(Byte[] buf, Int32 off, Int32 len) +73
   Thrift.Protocol.TBinaryProtocol.ReadI32() +53
   Thrift.Protocol.TBinaryProtocol.ReadMessageBegin() +79
   Apache.Cassandra.Client.recv_batch_mutate() +34
   Apache.Cassandra.Client.batch_mutate(Dictionary`2 mutation_map, ConsistencyLevel consistency_level) +24

But query is fine.

This is exception happened today, but in past month it worked fine.

Any idea?

New Post: Anyway to reset TTL for a column

paiwan — Thu, 29 Nov 2012 16:21:11 GMT

You just need to overwrite(insert) new column with the same name and new TTL.

New Post: Anyway to reset TTL for a column

dennisleng — Wed, 21 Nov 2012 02:07:28 GMT

Hi All

Just wondering anyway to reset TTL for a column as it could work the same way as asp.net sliding expiration?

Cheers

New Post: Exception 'Unable to read data from the transport connection:

vegah — Wed, 31 Oct 2012 07:55:53 GMT

I don't think I understand the DefaultClient.cs class. There is a do-while loop in the execute method, that retries MaximumRetries times - but as far as I can see, MaximumRetries is always set to 0, which means that it will never try more than one node. I though the idea was that if a node was down, it would try another one from the pool. I am probably misunderstanding something, because also if an exception is not marked as "ShouldRetry" the executioncounter is reset to 0 - which means that it will start from 0 again? What am I misunderstanding?

New Post: Exception 'Unable to read data from the transport connection:

vegah — Tue, 30 Oct 2012 13:09:13 GMT

I from time to time get this error when using the dev-webserver (included in vs) after a time of inactivity (hours, days). It does not seem to be able to recover without restarting the dev webserver.

Isn't this also a case where Aquiles should retry the connection or switch node? In DefaultCluster.cs I find a "catch (ExecutionBlockException ex)", but DefaultClient.buildException does not handle this exception. Not that I think it would matter, because as I said it does not seem to be able to recover without a restart. Any suggestions?

New Post: CompositeType Support

vegah — Tue, 30 Oct 2012 12:57:17 GMT

Added a patch that adds support for composite columns through a helper class. I have tested it, and it seems to work.

New Post: CompositeType Support

vegah — Fri, 19 Oct 2012 06:51:59 GMT

I am not sure if this is supported, I can not find anything in the code, or if it is planned. If there is no work done on it, I could take a shot at implementing it. I am not sure what would be the correct path, though. A CompositeType class, or if we are to follow the existing code maybe a ByteEncoderHelper<String,Int32>("Test",1)?

After reading http://www.jointhegrid.com/highperfcassandra/?p=152, I think it should be possible to fix either way.

New Post: Blacklisted node after Saving KeyCache

jaguarg — Fri, 12 Oct 2012 09:05:15 GMT

Thank you Javier,

yes we looked at the link above. And actually we had rowcache disabled. Now have disabled keycache but the same problem still occur.

We have found that reads are not affected and the exception is thrown on write operations only.

Any idea?

New Post: Blacklisted node after Saving KeyCache

javiercanillas — Thu, 11 Oct 2012 17:39:58 GMT

Have you read? http://www.datastax.com/docs/1.1/operations/tuning#cache-config

Can you please post the keyspace configuration? You might end up having to supress entirely key cache and row cache and see if that avoids the problem.

New Post: Blacklisted node after Saving KeyCache

jaguarg — Thu, 11 Oct 2012 13:10:31 GMT

Thank you Javier,

We are running Cassandra 1.1.2 but we had a the same problem on 0.7.4 before we upgraded.

Here is the description of the column family:

create column family ReportData
    with comparator = UTF8Type
    and memtable_throughput = 255
    and memtable_operations = 0.29
	and column_metadata = [
		{column_name : id			,	validation_class : UTF8Type,	index_name : id_idx		,	index_type : 0},		
		{column_name : eventid			,	validation_class : UTF8Type,	index_name : eventid_idx1		,	index_type : 0},
		{column_name : category,	validation_class : UTF8Type,	index_name : category_idx		,	index_type : 0},
		{column_name : samplerate,	validation_class : UTF8Type,	index_name : samplerate_idx0		,	index_type : 0},
		{column_name : unit,	validation_class : UTF8Type,	index_name : unit_idx		,	index_type : 0},
		{column_name : from,	validation_class : UTF8Type,	index_name : from_idx		,	index_type : 0},
		{column_name : to,	validation_class : UTF8Type,	index_name : to_idx		,	index_type : 0},
		{column_name : followerscount,	validation_class : UTF8Type,	index_name : followerscount_idx		,	index_type : 0},
		{column_name : screenname,	validation_class : UTF8Type,	index_name : screenname_idx		,	index_type : 0}

	] ;

Thank you

New Post: Blacklisted node after Saving KeyCache

jaguarg — Thu, 11 Oct 2012 13:10:30 GMT

Thank you Javier,

We are running Cassandra 1.1.2 but we had a the same problem on 0.7.4 before we upgraded.

Here is the description of the column family:

create column family ReportData
    with comparator = UTF8Type
    and memtable_throughput = 255
    and memtable_operations = 0.29
	and column_metadata = [
		{column_name : id			,	validation_class : UTF8Type,	index_name : id_idx		,	index_type : 0},		
		{column_name : eventid			,	validation_class : UTF8Type,	index_name : eventid_idx1		,	index_type : 0},
		{column_name : category,	validation_class : UTF8Type,	index_name : category_idx		,	index_type : 0},
		{column_name : samplerate,	validation_class : UTF8Type,	index_name : samplerate_idx0		,	index_type : 0},
		{column_name : unit,	validation_class : UTF8Type,	index_name : unit_idx		,	index_type : 0},
		{column_name : from,	validation_class : UTF8Type,	index_name : from_idx		,	index_type : 0},
		{column_name : to,	validation_class : UTF8Type,	index_name : to_idx		,	index_type : 0},
		{column_name : followerscount,	validation_class : UTF8Type,	index_name : followerscount_idx		,	index_type : 0},
		{column_name : screenname,	validation_class : UTF8Type,	index_name : screenname_idx		,	index_type : 0}

	] ;

Thank you

New Post: Blacklisted node after Saving KeyCache

javiercanillas — Tue, 09 Oct 2012 14:20:19 GMT

It seems your problem is within Cassandra itself. By doing a node restart, and since you only have one, All connections inside Aquiles are broken. That's how Aquiles ends up saying that.

Besides asking in the Cassandra mailing list, can you post what version of Cassandra are you using, and the CQL of your keyspace and column family? I can try to help you that way.

New Post: Is UnixHelper.UnixTimestamp accurate?

javiercanillas — Tue, 09 Oct 2012 14:18:10 GMT

Sure, I will take a look at it. It's weird you are having divived by zero errors.

New Post: Blacklisted node after Saving KeyCache

jaguarg — Tue, 09 Oct 2012 07:13:36 GMT

I have an interesting issue with cassandra and aquiles that happens when cassandra Saves half a dozen KeyCaches.

In fact, when looking at the log, I can see that after

INFO ... Saved KeyCache (.... )

my app (that uses aquiles) is not longer able to access our cassandra node (we run a single node). When I restart it, I get a message: "All endpoints ['127.0.0.1:9160-6000'] are blacklisted, cluster is down?"

The only option I am left with is to restart Cassandra. I have had to do this every couple of days.

Any idea?

New Post: Is UnixHelper.UnixTimestamp accurate?

paiwan — Mon, 08 Oct 2012 08:20:09 GMT

I am using version 1.0.0.1 running 3 nodes of Cassandra on Windows 2008 Hyper-V

Has anyone tried to compare the results between UnixHelper.UnixTimestamp and adding to Cassandra directly?

I've found the problem, the timestamps are difference.

I am not sure if the problem happen only me but after looking into UnixHelper.UnixTimestamp

The following is the details how I fix.

I changed from..

static readonly DateTime unitStartDateTime = new DateTime(1970, 1, 1).ToUniversalTime();

to..

static readonly DateTime unitStartDateTime = new DateTime(1970, 1, 1, 0, 0, 0, 0, DateTimeKind.Utc);

after changing it seems I got better result but...

Another problem is "Devide by zero" from class DateTimePrecise.UtcNow(not everytime but more than 10 times a day)

return immutable._t_base.AddTicks(((
    s - immutable._s_observed) * _clockTickFrequency) / (
    immutable._stopWatchFrequency));

So I decided to modify the whole branch of code for UnixHelper.UnixTimestamp as follow:

using System.Diagnostics;
namespace xxx
{
    public class UnixHelper
    {
        static readonly DateTime unitStartDateTime = new DateTime(1970, 1, 1, 0, 0, 0, 0, DateTimeKind.Utc);

        public static long UnixTimestamp
        {
            get
            {
                TimeSpan difference = HiResDateTime.UtcNow - unitStartDateTime;
                return difference.Ticks / 10;
            }
        }
    }


    public class HiResDateTime
    {
        private static DateTime _startTime;
        private static Stopwatch _stopWatch = null;
        private static TimeSpan _maxIdle =
            TimeSpan.FromSeconds(10);

        public static DateTime UtcNow
        {
            get
            {
                if ((_stopWatch == null) ||
                    (_startTime.Add(_maxIdle) < DateTime.UtcNow))
                {
                    Reset();
                }
                return _startTime.AddTicks(_stopWatch.Elapsed.Ticks);
            }
        }

        private static void Reset()
        {
            _startTime = DateTime.UtcNow;
            _stopWatch = Stopwatch.StartNew();
        }
    }
}

I am not sure if the code is better than the current version from Aquiles but for me it's perfect. no problem at all after modified the code.

If the above code is the correct way to handle UnixTimestamp, please include in the next version.

New Post: AquilesHelper.RetrieveCluster

javiercanillas — Fri, 28 Sep 2012 14:10:38 GMT

Amit,

I'm happy to hear your problem is solved. Even though, I will double-check that ambiguity and see if I can get rid of it to avoid future mistakes.

New Post: AquilesHelper.RetrieveCluster

AmitShankar — Fri, 28 Sep 2012 11:42:11 GMT

Hello,

It was my bad. But a finding though, AquilesHelper is present in both Cassandra10 and Aquiles.Core. The one in Cassandra10 is to be used. I had included both namespaces and hence the ambiguity.

Thanks, Amit

New Post: AquilesHelper.RetrieveCluster

AmitShankar — Thu, 27 Sep 2012 14:45:19 GMT

Hello,

I am using Aquiles 1.0.0.1 to code a .Net web service as a layer over Cassandra. I am simply trying to insert records in my Cassandra DB. I did it before with 0.8 and below stated statement worked.

ICluster ObjCluster = AquilesHelper.RetrieveCluster("Test Cluster");

Now when I am trying to do the same with 1.0, its not allowing me to use this as above. The error says that i need to instantiate an object and RetrieveCluster is not directly available.

On defining something like below and using my AqlHelper object, it seems that I have to define much more again.

AquilesHelper AqlHelper = new AquilesHelper(AbstractAquilesClusterBuilder,"aquilesConfiguration");

it seems that i have to predefine my cluster again using AbstractAquilesClusterBuilder when I have already defined the details in my web.config file. AbstractAquilesClusterBuilder is again an abstract class with logger as another interface in its parameter while building it. Also ClusterConfig is to be defined again.

Is there a simpler way of doing it or using what I have already defined in my web.config file.

Thanks, Amit

New Post: Dictionary in CqlMetadata.cs

javiercanillas — Fri, 21 Sep 2012 00:43:58 GMT

Thanks, I will give it a look and let you know.

New Post: Recover from All endpoints are blacklisted exception

javiercanillas — Fri, 21 Sep 2012 00:43:34 GMT

From what I know, nodetool is only avaiable through the command line and, not sure, through some administrative application distributed by DataStax. Although, Nodetool is sending its command through a JMX interface, so if you see it's code you may achieve the same some way.

About killing your cassandra nodes, the only way through Linux is by doing a "kill" command and then executing the process again. On windows, you may need to do a service restart. Nevertheless, I wouldn't sugguest going on that direction, it is much more likely you are having a problem, and trying to fix that problem will assure system stability.