Classified as: {Hitachi Rail – Internal}
Hi Matt,
I'd be surprised if this was the case as swapping between between Linux and
Windows VMs consistently shows the pattern of not keeping up/keeping up with
the instance type the same. I also think that the VM itself is over provisioned
to reduce the probability of getting a low quality VM. Just to be clear we're
not using any t instances that do CPU bursting. We are using an m7i.xlarge
instance which has 4 vCPU and 16GB RAM. However, this is definitely something
we can check and could also explore a dedicated host temporarily or putting in
a higher capability instance to see if that changes anything.
Rob
{Hitachi Rail – Internal}
-----Original Message-----
From: Matt Pavlovich <[email protected]>
Sent: 12 February 2026 16:29
To: [email protected]
Cc: KIRKBRIDE Robert <[email protected]>
Subject: Re: ActiveMQ Classic Linux Broker Performance Issues
[You don't often get email from [email protected]. Learn why this is
important at https://aka.ms/LearnAboutSenderIdentification ]
Benchmark the VM by itself using OS benchmark tools to rule out the host
itself. You could be running into low-quality Linux VM running on an
over-subscribed physical host.
-Matt
> On Feb 11, 2026, at 12:31 PM, Rob Kirkbride <[email protected]> wrote:
>
> Yes the same in both cases I believe.
> We are in an Aws environment so we've tried both Amazon Linux and
> Ubuntu Linux and applied all available patches.
> We've also tried increasing the iops on the SSD storage, though
> there's no evidence that it's a bottleneck.
>
> On Wed, 11 Feb 2026, 18:27 Jean-Baptiste Onofré, <[email protected]> wrote:
>
>> Hi
>>
>> Are you using the same activemq.xml ? Same JVM configuration ?
>>
>> We would need more details to investigate.
>>
>> Regards
>> JB
>>
>> On Wed, Feb 11, 2026 at 6:02 PM KIRKBRIDE Robert via users <
>> [email protected]> wrote:
>>
>>> Classified as: {Hitachi Rail - Public}
>>>
>>> We seem to have hit an issue with Linux Broker vs Windows Broker
>>> performance that we can't understand or resolve.
>>> We are using v6.2.0 of ActiveMQ Classic. We have a service that is
>> pushing
>>> messages to a virtual topic (VirtualTopic.Name), typically around
>>> 20-30 bytes and these are being published at a frequency of around
>>> 130
>> messages /
>>> second. We then have two services consuming from separate virtual
>>> topic queues (Consumer.service.VirtualTopic.Name). On a Windows
>>> broker we are seeing ActiveMQ easily deal with this, with low CPU and
>>> memory usage.
>> With
>>> a Linux broker the messages are being consumed but seemingly not
>>> fast enough as the queues build up over time (probably running about
>>> half the speed required) The specifications of the Linux and Windows
>>> VMs are the same and in both cases, memory and CPU is low.
>>> The clients are C# .NET services. Nothing else about the environment
>>> changes. Any ideas as how we can resolve this, anything to try?
>>> Thanks a lot in advance.
>>>
>>>
>>> {Hitachi Rail - Public}
>>>
>>
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
For further information, visit: https://activemq.apache.org/contact
