You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@artemis.apache.org by Thorsten Meinl <th...@knime.com> on 2026/09/22 10:52:06 UTC
Slow connects after updating to 2.57.0
Hello everyone,
After updating our Artemis broker to 2.57.0 we are experiencing high
latencies when clients connect to the broker via the CORE protocol. The
clients themselves have not been changed. They are currently using
version 2.52.0 of the JMS client lib (pulled in by a dependency chain).
Client are all authenticated and connect via
jakarta.jms.ConnectionFactory.createContext().
After reading the release notes, we believe it has something to do with
the changed Core protocol connect handshake. However, it is also
mentioned that the change is "backward-compatible" and therefore we did
not pay much attention. So a couple of questions:
- Could this broker change be responsible for the increased connection
times?
- If so, how can we fix it without changing all clients?
- Is this effect expected/intended?
Please let me know, if you need more information.
Thanks and regards,
Thorsten
--
Dr.-Ing. Thorsten Meinl
KNIME AG
Talacker 50
8001 Zurich, Switzerland
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
For additional commands, e-mail: users-help@artemis.apache.org
Re: Slow connects after updating to 2.57.0
Posted by Thorsten Meinl <th...@knime.com>.
2.56.0 works as well.
Justin Bertram <jb...@apache.org> schrieb am Di., 22. Sept. 2026, 18:27:
> Have you tested with 2.56.0? Knowing the results of that test would
> help isolate the issue assuming the issue is actually on the broker.
> It could have been something in 2.56.0 rather than 2.57.0.
>
>
> Justin
>
> On Tue, Sep 22, 2026 at 10:30 AM Thorsten Meinl
> <th...@knime.com> wrote:
> >
> > Am Dienstag, dem 22.09.2026 um 08:43 -0500 schrieb Justin Bertram:
> > > Did you move the broker from 2.56.0 to 2.57.0?
> > We updated from 2.55.0 to 2.57.0.
> >
> > Attached are two logs from an affected service. The only difference is
> > the broker version it connects to (2.55 vs 2.57). As you can see, with
> > 2.57 is takes considerably longer between the two INFO message (~1s vs.
> > ~9s). In the 2.57 case I see remote destinations such as
> > localhost/127.0.0.1:61616 or /0.0.0.0:61616 being tried. Which
> > obviously doesn't work and seems to me to be the cause for the delays.
> >
> > Maybe this information helps.
> >
> > Thanks,
> >
> > Thorsten
> >
> > > On Tue, Sep 22, 2026 at 5:52 AM Thorsten Meinl
> > > <th...@knime.com> wrote:
> > > >
> > > > Hello everyone,
> > > >
> > > > After updating our Artemis broker to 2.57.0 we are experiencing
> > > > high
> > > > latencies when clients connect to the broker via the CORE protocol.
> > > > The
> > > > clients themselves have not been changed. They are currently using
> > > > version 2.52.0 of the JMS client lib (pulled in by a dependency
> > > > chain).
> > > > Client are all authenticated and connect via
> > > > jakarta.jms.ConnectionFactory.createContext().
> > > >
> > > > After reading the release notes, we believe it has something to do
> > > > with
> > > > the changed Core protocol connect handshake. However, it is also
> > > > mentioned that the change is "backward-compatible" and therefore we
> > > > did
> > > > not pay much attention. So a couple of questions:
> > > >
> > > > - Could this broker change be responsible for the increased
> > > > connection
> > > > times?
> > > > - If so, how can we fix it without changing all clients?
> > > > - Is this effect expected/intended?
> > > >
> > > > Please let me know, if you need more information.
> > > >
> > > > Thanks and regards,
> > > >
> > > > Thorsten
> > > >
> > > > --
> > > > Dr.-Ing. Thorsten Meinl
> > > > KNIME AG
> > > > Talacker 50
> > > > 8001 Zurich, Switzerland
> > > >
> > > >
> > > > -------------------------------------------------------------------
> > > > --
> > > > To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> > > > For additional commands, e-mail: users-help@artemis.apache.org
> > > >
> > >
> > > ---------------------------------------------------------------------
> > > To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> > > For additional commands, e-mail: users-help@artemis.apache.org
> >
> > --
> > Dr.-Ing. Thorsten Meinl
> > KNIME AG
> > Talacker 50
> > 8001 Zurich, Switzerland
> >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> > For additional commands, e-mail: users-help@artemis.apache.org
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> For additional commands, e-mail: users-help@artemis.apache.org
>
>
Re: Slow connects after updating to 2.57.0
Posted by Thorsten Meinl <th...@knime.com>.
Am Freitag, dem 25.09.2026 um 16:19 -0400 schrieb Clebert Suconic:
> Can you upgrade your clients. Would still be an issue ?
Unfortunately this is not possible yet therefore I cannot tell.
> On Wed, Sep 23, 2026 at 7:25 AM Thorsten Meinl
> <th...@knime.com> wrote:
> > I threw some logs and the 2.57.0 source at Claude and it came up
> > with a
> > plausible cause. Please check whether it's valid and deserves a bug
> > fix.
> >
> > ---------
> > It starts with CoreProtocolManager.java:370:
> >
> > if (server.getSecurityStore().isSecurityEnabled() &&
> > isCoreConnectionSecurityEnabled() && !rc.isAuthenticated()) {
> > if (((SubscribeClusterTopologyUpdatesMessage)
> > packet).isClusterConnection()) {
> >
> > ActiveMQServerLogger.LOGGER.unauthenticatedClusterTopologySubscript
> > ionR
> > equest(rc.getRemoteAddress());
> > }
> > sendPreAuthTopology();
> > } else {
> > sendPostAuthTopology();
> > }
> >
> > A Core client subscribes to topology updates on channel 0 before it
> > authenticates. 2.57 hardened this so the real topology is no longer
> > disclosed to an unauthenticated caller. But the client blocks
> > waiting
> > for a topology message before it proceeds to CREATE_SESSION, so the
> > broker has to send something. That something is
> > sendPreAuthTopology() —
> > a placeholder member under node ID 00000000-0000-0000-0000-
> > 000000000000.
> >
> > Here's where your two bad addresses (0.0.0.0 and localhost) come
> > from
> > (CoreProtocolManager.java:387-405):
> >
> > TransportConfiguration connectorConfig =
> > rc.getTransportConnection().getConnectorConfig();
> > // ... copies HOST_PROP_NAME and PORT_PROP_NAME out of it ...
> > TransportConfiguration liveConnector = new
> > TransportConfiguration(factoryClassName, connectorParams);
> > Pair<TransportConfiguration, TransportConfiguration> preAuthConfig
> > =
> > ...
> > new Pair<>(liveConnector, new
> > TransportConfiguration(factoryClassName, null,
> > Topology.PRE_AUTH_CONNECTOR_NAME));
> >
> >
> > The primary comes from NettyConnection.getConnectorConfig(), which
> > is
> > new TransportConfiguration(NettyConnectorFactory.class.getName(),
> > this.configuration) — the acceptor's own config map. Your acceptor
> > is
> > tcp://0.0.0.0:61616, so the advertised host is 0.0.0.0.
> > The backup is constructed with null params, so Netty applies its
> > default host: localhost.
> >
> > That is exactly the pair in your client log. And the Notifying
> > 00000000-0000-0000-0000-000000000000 going down line is the cleanup
> > nodeDown the broker sends once authentication completes.
> >
> > The 2.57 client knows to ignore the placeholder
> > (ClientSessionFactoryImpl.java:1557):
> >
> > java
> > // Skip notifications for PRE_AUTH_NODE_ID
> > if (!Topology.PRE_AUTH_NODE_ID.equals(nodeID)) {
> > ...
> > serverLocator.notifyNodeUp(uniqueEventID, nodeID,
> > backupGroupName,
> > scaleDownGroupName, connectorPair, isLast);
> > }
> >
> > Your client is older than 2.57. It ingests the placeholder as a
> > genuine
> > topology member and dutifully tries 0.0.0.0, then localhost, then
> > backs
> > off 2 seconds.
> >
> > The comment above the method states the intent plainly: the current
> > connector is sent so that older clients wouldn't fail and would
> > have
> > something they can connect to. The compatibility assumption breaks
> > when
> > the acceptor binds to a wildcard, because 0.0.0.0 is not something
> > anyone can connect to — and the localhost fallback is never right
> > from
> > a remote client.
> > ---------
> >
> > Thanks,
> >
> > Thorsten
> >
> > Am Dienstag, dem 22.09.2026 um 11:27 -0500 schrieb Justin Bertram:
> > > Have you tested with 2.56.0? Knowing the results of that test
> > > would
> > > help isolate the issue assuming the issue is actually on the
> > > broker.
> > > It could have been something in 2.56.0 rather than 2.57.0.
> > >
> > >
> > > Justin
> > >
> > > On Tue, Sep 22, 2026 at 10:30 AM Thorsten Meinl
> > > <th...@knime.com> wrote:
> > > >
> > > > Am Dienstag, dem 22.09.2026 um 08:43 -0500 schrieb Justin
> > > > Bertram:
> > > > > Did you move the broker from 2.56.0 to 2.57.0?
> > > > We updated from 2.55.0 to 2.57.0.
> > > >
> > > > Attached are two logs from an affected service. The only
> > > > difference
> > > > is
> > > > the broker version it connects to (2.55 vs 2.57). As you can
> > > > see,
> > > > with
> > > > 2.57 is takes considerably longer between the two INFO message
> > > > (~1s
> > > > vs.
> > > > ~9s). In the 2.57 case I see remote destinations such as
> > > > localhost/127.0.0.1:61616 or /0.0.0.0:61616 being tried. Which
> > > > obviously doesn't work and seems to me to be the cause for the
> > > > delays.
> > > >
> > > > Maybe this information helps.
> > > >
> > > > Thanks,
> > > >
> > > > Thorsten
> > > >
> > > > > On Tue, Sep 22, 2026 at 5:52 AM Thorsten Meinl
> > > > > <th...@knime.com> wrote:
> > > > > >
> > > > > > Hello everyone,
> > > > > >
> > > > > > After updating our Artemis broker to 2.57.0 we are
> > > > > > experiencing
> > > > > > high
> > > > > > latencies when clients connect to the broker via the CORE
> > > > > > protocol.
> > > > > > The
> > > > > > clients themselves have not been changed. They are
> > > > > > currently
> > > > > > using
> > > > > > version 2.52.0 of the JMS client lib (pulled in by a
> > > > > > dependency
> > > > > > chain).
> > > > > > Client are all authenticated and connect via
> > > > > > jakarta.jms.ConnectionFactory.createContext().
> > > > > >
> > > > > > After reading the release notes, we believe it has
> > > > > > something to
> > > > > > do
> > > > > > with
> > > > > > the changed Core protocol connect handshake. However, it is
> > > > > > also
> > > > > > mentioned that the change is "backward-compatible" and
> > > > > > therefore we
> > > > > > did
> > > > > > not pay much attention. So a couple of questions:
> > > > > >
> > > > > > - Could this broker change be responsible for the increased
> > > > > > connection
> > > > > > times?
> > > > > > - If so, how can we fix it without changing all clients?
> > > > > > - Is this effect expected/intended?
> > > > > >
> > > > > > Please let me know, if you need more information.
> > > > > >
> > > > > > Thanks and regards,
> > > > > >
> > > > > > Thorsten
> > > > > >
> > > > > > --
> > > > > > Dr.-Ing. Thorsten Meinl
> > > > > > KNIME AG
> > > > > > Talacker 50
> > > > > > 8001 Zurich, Switzerland
> > > > > >
> > > > > >
> > > > > > -----------------------------------------------------------
> > > > > > ----
> > > > > > ----
> > > > > > --
> > > > > > To unsubscribe, e-mail:
> > > > > > users-unsubscribe@artemis.apache.org
> > > > > > For additional commands, e-mail:
> > > > > > users-help@artemis.apache.org
> > > > > >
> > > > >
> > > > > -------------------------------------------------------------
> > > > > ----
> > > > > ----
> > > > > To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> > > > > For additional commands, e-mail:
> > > > > users-help@artemis.apache.org
> > > >
> > > > --
> > > > Dr.-Ing. Thorsten Meinl
> > > > KNIME AG
> > > > Talacker 50
> > > > 8001 Zurich, Switzerland
> > > >
> > > >
> > > > ---------------------------------------------------------------
> > > > ----
> > > > --
> > > > To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> > > > For additional commands, e-mail: users-help@artemis.apache.org
> > >
> > > -----------------------------------------------------------------
> > > ----
> > > To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> > > For additional commands, e-mail: users-help@artemis.apache.org
> >
--
Dr.-Ing. Thorsten Meinl
KNIME AG
Talacker 50
8001 Zurich, Switzerland
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
For additional commands, e-mail: users-help@artemis.apache.org
Re: Slow connects after updating to 2.57.0
Posted by Clebert Suconic <cl...@gmail.com>.
Can you upgrade your clients. Would still be an issue ?
Clebert Suconic
On Wed, Sep 23, 2026 at 7:25 AM Thorsten Meinl <th...@knime.com>
wrote:
> I threw some logs and the 2.57.0 source at Claude and it came up with a
> plausible cause. Please check whether it's valid and deserves a bug
> fix.
>
> ---------
> It starts with CoreProtocolManager.java:370:
>
> if (server.getSecurityStore().isSecurityEnabled() &&
> isCoreConnectionSecurityEnabled() && !rc.isAuthenticated()) {
> if (((SubscribeClusterTopologyUpdatesMessage)
> packet).isClusterConnection()) {
>
> ActiveMQServerLogger.LOGGER.unauthenticatedClusterTopologySubscriptionR
> equest(rc.getRemoteAddress());
> }
> sendPreAuthTopology();
> } else {
> sendPostAuthTopology();
> }
>
> A Core client subscribes to topology updates on channel 0 before it
> authenticates. 2.57 hardened this so the real topology is no longer
> disclosed to an unauthenticated caller. But the client blocks waiting
> for a topology message before it proceeds to CREATE_SESSION, so the
> broker has to send something. That something is sendPreAuthTopology() —
> a placeholder member under node ID 00000000-0000-0000-0000-
> 000000000000.
>
> Here's where your two bad addresses (0.0.0.0 and localhost) come from
> (CoreProtocolManager.java:387-405):
>
> TransportConfiguration connectorConfig =
> rc.getTransportConnection().getConnectorConfig();
> // ... copies HOST_PROP_NAME and PORT_PROP_NAME out of it ...
> TransportConfiguration liveConnector = new
> TransportConfiguration(factoryClassName, connectorParams);
> Pair<TransportConfiguration, TransportConfiguration> preAuthConfig =
> ...
> new Pair<>(liveConnector, new
> TransportConfiguration(factoryClassName, null,
> Topology.PRE_AUTH_CONNECTOR_NAME));
>
>
> The primary comes from NettyConnection.getConnectorConfig(), which is
> new TransportConfiguration(NettyConnectorFactory.class.getName(),
> this.configuration) — the acceptor's own config map. Your acceptor is
> tcp://0.0.0.0:61616, so the advertised host is 0.0.0.0.
> The backup is constructed with null params, so Netty applies its
> default host: localhost.
>
> That is exactly the pair in your client log. And the Notifying
> 00000000-0000-0000-0000-000000000000 going down line is the cleanup
> nodeDown the broker sends once authentication completes.
>
> The 2.57 client knows to ignore the placeholder
> (ClientSessionFactoryImpl.java:1557):
>
> java
> // Skip notifications for PRE_AUTH_NODE_ID
> if (!Topology.PRE_AUTH_NODE_ID.equals(nodeID)) {
> ...
> serverLocator.notifyNodeUp(uniqueEventID, nodeID, backupGroupName,
> scaleDownGroupName, connectorPair, isLast);
> }
>
> Your client is older than 2.57. It ingests the placeholder as a genuine
> topology member and dutifully tries 0.0.0.0, then localhost, then backs
> off 2 seconds.
>
> The comment above the method states the intent plainly: the current
> connector is sent so that older clients wouldn't fail and would have
> something they can connect to. The compatibility assumption breaks when
> the acceptor binds to a wildcard, because 0.0.0.0 is not something
> anyone can connect to — and the localhost fallback is never right from
> a remote client.
> ---------
>
> Thanks,
>
> Thorsten
>
> Am Dienstag, dem 22.09.2026 um 11:27 -0500 schrieb Justin Bertram:
> > Have you tested with 2.56.0? Knowing the results of that test would
> > help isolate the issue assuming the issue is actually on the broker.
> > It could have been something in 2.56.0 rather than 2.57.0.
> >
> >
> > Justin
> >
> > On Tue, Sep 22, 2026 at 10:30 AM Thorsten Meinl
> > <th...@knime.com> wrote:
> > >
> > > Am Dienstag, dem 22.09.2026 um 08:43 -0500 schrieb Justin Bertram:
> > > > Did you move the broker from 2.56.0 to 2.57.0?
> > > We updated from 2.55.0 to 2.57.0.
> > >
> > > Attached are two logs from an affected service. The only difference
> > > is
> > > the broker version it connects to (2.55 vs 2.57). As you can see,
> > > with
> > > 2.57 is takes considerably longer between the two INFO message (~1s
> > > vs.
> > > ~9s). In the 2.57 case I see remote destinations such as
> > > localhost/127.0.0.1:61616 or /0.0.0.0:61616 being tried. Which
> > > obviously doesn't work and seems to me to be the cause for the
> > > delays.
> > >
> > > Maybe this information helps.
> > >
> > > Thanks,
> > >
> > > Thorsten
> > >
> > > > On Tue, Sep 22, 2026 at 5:52 AM Thorsten Meinl
> > > > <th...@knime.com> wrote:
> > > > >
> > > > > Hello everyone,
> > > > >
> > > > > After updating our Artemis broker to 2.57.0 we are experiencing
> > > > > high
> > > > > latencies when clients connect to the broker via the CORE
> > > > > protocol.
> > > > > The
> > > > > clients themselves have not been changed. They are currently
> > > > > using
> > > > > version 2.52.0 of the JMS client lib (pulled in by a dependency
> > > > > chain).
> > > > > Client are all authenticated and connect via
> > > > > jakarta.jms.ConnectionFactory.createContext().
> > > > >
> > > > > After reading the release notes, we believe it has something to
> > > > > do
> > > > > with
> > > > > the changed Core protocol connect handshake. However, it is
> > > > > also
> > > > > mentioned that the change is "backward-compatible" and
> > > > > therefore we
> > > > > did
> > > > > not pay much attention. So a couple of questions:
> > > > >
> > > > > - Could this broker change be responsible for the increased
> > > > > connection
> > > > > times?
> > > > > - If so, how can we fix it without changing all clients?
> > > > > - Is this effect expected/intended?
> > > > >
> > > > > Please let me know, if you need more information.
> > > > >
> > > > > Thanks and regards,
> > > > >
> > > > > Thorsten
> > > > >
> > > > > --
> > > > > Dr.-Ing. Thorsten Meinl
> > > > > KNIME AG
> > > > > Talacker 50
> > > > > 8001 Zurich, Switzerland
> > > > >
> > > > >
> > > > > ---------------------------------------------------------------
> > > > > ----
> > > > > --
> > > > > To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> > > > > For additional commands, e-mail: users-help@artemis.apache.org
> > > > >
> > > >
> > > > -----------------------------------------------------------------
> > > > ----
> > > > To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> > > > For additional commands, e-mail: users-help@artemis.apache.org
> > >
> > > --
> > > Dr.-Ing. Thorsten Meinl
> > > KNIME AG
> > > Talacker 50
> > > 8001 Zurich, Switzerland
> > >
> > >
> > > -------------------------------------------------------------------
> > > --
> > > To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> > > For additional commands, e-mail: users-help@artemis.apache.org
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> > For additional commands, e-mail: users-help@artemis.apache.org
>
> --
> Dr.-Ing. Thorsten Meinl
> KNIME AG
> Talacker 50
> 8001 Zurich, Switzerland
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> For additional commands, e-mail: users-help@artemis.apache.org
>
>
Re: Slow connects after updating to 2.57.0
Posted by Thorsten Meinl <th...@knime.com>.
I threw some logs and the 2.57.0 source at Claude and it came up with a
plausible cause. Please check whether it's valid and deserves a bug
fix.
---------
It starts with CoreProtocolManager.java:370:
if (server.getSecurityStore().isSecurityEnabled() &&
isCoreConnectionSecurityEnabled() && !rc.isAuthenticated()) {
if (((SubscribeClusterTopologyUpdatesMessage)
packet).isClusterConnection()) {
ActiveMQServerLogger.LOGGER.unauthenticatedClusterTopologySubscriptionR
equest(rc.getRemoteAddress());
}
sendPreAuthTopology();
} else {
sendPostAuthTopology();
}
A Core client subscribes to topology updates on channel 0 before it
authenticates. 2.57 hardened this so the real topology is no longer
disclosed to an unauthenticated caller. But the client blocks waiting
for a topology message before it proceeds to CREATE_SESSION, so the
broker has to send something. That something is sendPreAuthTopology() —
a placeholder member under node ID 00000000-0000-0000-0000-
000000000000.
Here's where your two bad addresses (0.0.0.0 and localhost) come from
(CoreProtocolManager.java:387-405):
TransportConfiguration connectorConfig =
rc.getTransportConnection().getConnectorConfig();
// ... copies HOST_PROP_NAME and PORT_PROP_NAME out of it ...
TransportConfiguration liveConnector = new
TransportConfiguration(factoryClassName, connectorParams);
Pair<TransportConfiguration, TransportConfiguration> preAuthConfig =
...
new Pair<>(liveConnector, new
TransportConfiguration(factoryClassName, null,
Topology.PRE_AUTH_CONNECTOR_NAME));
The primary comes from NettyConnection.getConnectorConfig(), which is
new TransportConfiguration(NettyConnectorFactory.class.getName(),
this.configuration) — the acceptor's own config map. Your acceptor is
tcp://0.0.0.0:61616, so the advertised host is 0.0.0.0.
The backup is constructed with null params, so Netty applies its
default host: localhost.
That is exactly the pair in your client log. And the Notifying
00000000-0000-0000-0000-000000000000 going down line is the cleanup
nodeDown the broker sends once authentication completes.
The 2.57 client knows to ignore the placeholder
(ClientSessionFactoryImpl.java:1557):
java
// Skip notifications for PRE_AUTH_NODE_ID
if (!Topology.PRE_AUTH_NODE_ID.equals(nodeID)) {
...
serverLocator.notifyNodeUp(uniqueEventID, nodeID, backupGroupName,
scaleDownGroupName, connectorPair, isLast);
}
Your client is older than 2.57. It ingests the placeholder as a genuine
topology member and dutifully tries 0.0.0.0, then localhost, then backs
off 2 seconds.
The comment above the method states the intent plainly: the current
connector is sent so that older clients wouldn't fail and would have
something they can connect to. The compatibility assumption breaks when
the acceptor binds to a wildcard, because 0.0.0.0 is not something
anyone can connect to — and the localhost fallback is never right from
a remote client.
---------
Thanks,
Thorsten
Am Dienstag, dem 22.09.2026 um 11:27 -0500 schrieb Justin Bertram:
> Have you tested with 2.56.0? Knowing the results of that test would
> help isolate the issue assuming the issue is actually on the broker.
> It could have been something in 2.56.0 rather than 2.57.0.
>
>
> Justin
>
> On Tue, Sep 22, 2026 at 10:30 AM Thorsten Meinl
> <th...@knime.com> wrote:
> >
> > Am Dienstag, dem 22.09.2026 um 08:43 -0500 schrieb Justin Bertram:
> > > Did you move the broker from 2.56.0 to 2.57.0?
> > We updated from 2.55.0 to 2.57.0.
> >
> > Attached are two logs from an affected service. The only difference
> > is
> > the broker version it connects to (2.55 vs 2.57). As you can see,
> > with
> > 2.57 is takes considerably longer between the two INFO message (~1s
> > vs.
> > ~9s). In the 2.57 case I see remote destinations such as
> > localhost/127.0.0.1:61616 or /0.0.0.0:61616 being tried. Which
> > obviously doesn't work and seems to me to be the cause for the
> > delays.
> >
> > Maybe this information helps.
> >
> > Thanks,
> >
> > Thorsten
> >
> > > On Tue, Sep 22, 2026 at 5:52 AM Thorsten Meinl
> > > <th...@knime.com> wrote:
> > > >
> > > > Hello everyone,
> > > >
> > > > After updating our Artemis broker to 2.57.0 we are experiencing
> > > > high
> > > > latencies when clients connect to the broker via the CORE
> > > > protocol.
> > > > The
> > > > clients themselves have not been changed. They are currently
> > > > using
> > > > version 2.52.0 of the JMS client lib (pulled in by a dependency
> > > > chain).
> > > > Client are all authenticated and connect via
> > > > jakarta.jms.ConnectionFactory.createContext().
> > > >
> > > > After reading the release notes, we believe it has something to
> > > > do
> > > > with
> > > > the changed Core protocol connect handshake. However, it is
> > > > also
> > > > mentioned that the change is "backward-compatible" and
> > > > therefore we
> > > > did
> > > > not pay much attention. So a couple of questions:
> > > >
> > > > - Could this broker change be responsible for the increased
> > > > connection
> > > > times?
> > > > - If so, how can we fix it without changing all clients?
> > > > - Is this effect expected/intended?
> > > >
> > > > Please let me know, if you need more information.
> > > >
> > > > Thanks and regards,
> > > >
> > > > Thorsten
> > > >
> > > > --
> > > > Dr.-Ing. Thorsten Meinl
> > > > KNIME AG
> > > > Talacker 50
> > > > 8001 Zurich, Switzerland
> > > >
> > > >
> > > > ---------------------------------------------------------------
> > > > ----
> > > > --
> > > > To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> > > > For additional commands, e-mail: users-help@artemis.apache.org
> > > >
> > >
> > > -----------------------------------------------------------------
> > > ----
> > > To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> > > For additional commands, e-mail: users-help@artemis.apache.org
> >
> > --
> > Dr.-Ing. Thorsten Meinl
> > KNIME AG
> > Talacker 50
> > 8001 Zurich, Switzerland
> >
> >
> > -------------------------------------------------------------------
> > --
> > To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> > For additional commands, e-mail: users-help@artemis.apache.org
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> For additional commands, e-mail: users-help@artemis.apache.org
--
Dr.-Ing. Thorsten Meinl
KNIME AG
Talacker 50
8001 Zurich, Switzerland
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
For additional commands, e-mail: users-help@artemis.apache.org
Re: Slow connects after updating to 2.57.0
Posted by Justin Bertram <jb...@apache.org>.
Have you tested with 2.56.0? Knowing the results of that test would
help isolate the issue assuming the issue is actually on the broker.
It could have been something in 2.56.0 rather than 2.57.0.
Justin
On Tue, Sep 22, 2026 at 10:30 AM Thorsten Meinl
<th...@knime.com> wrote:
>
> Am Dienstag, dem 22.09.2026 um 08:43 -0500 schrieb Justin Bertram:
> > Did you move the broker from 2.56.0 to 2.57.0?
> We updated from 2.55.0 to 2.57.0.
>
> Attached are two logs from an affected service. The only difference is
> the broker version it connects to (2.55 vs 2.57). As you can see, with
> 2.57 is takes considerably longer between the two INFO message (~1s vs.
> ~9s). In the 2.57 case I see remote destinations such as
> localhost/127.0.0.1:61616 or /0.0.0.0:61616 being tried. Which
> obviously doesn't work and seems to me to be the cause for the delays.
>
> Maybe this information helps.
>
> Thanks,
>
> Thorsten
>
> > On Tue, Sep 22, 2026 at 5:52 AM Thorsten Meinl
> > <th...@knime.com> wrote:
> > >
> > > Hello everyone,
> > >
> > > After updating our Artemis broker to 2.57.0 we are experiencing
> > > high
> > > latencies when clients connect to the broker via the CORE protocol.
> > > The
> > > clients themselves have not been changed. They are currently using
> > > version 2.52.0 of the JMS client lib (pulled in by a dependency
> > > chain).
> > > Client are all authenticated and connect via
> > > jakarta.jms.ConnectionFactory.createContext().
> > >
> > > After reading the release notes, we believe it has something to do
> > > with
> > > the changed Core protocol connect handshake. However, it is also
> > > mentioned that the change is "backward-compatible" and therefore we
> > > did
> > > not pay much attention. So a couple of questions:
> > >
> > > - Could this broker change be responsible for the increased
> > > connection
> > > times?
> > > - If so, how can we fix it without changing all clients?
> > > - Is this effect expected/intended?
> > >
> > > Please let me know, if you need more information.
> > >
> > > Thanks and regards,
> > >
> > > Thorsten
> > >
> > > --
> > > Dr.-Ing. Thorsten Meinl
> > > KNIME AG
> > > Talacker 50
> > > 8001 Zurich, Switzerland
> > >
> > >
> > > -------------------------------------------------------------------
> > > --
> > > To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> > > For additional commands, e-mail: users-help@artemis.apache.org
> > >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> > For additional commands, e-mail: users-help@artemis.apache.org
>
> --
> Dr.-Ing. Thorsten Meinl
> KNIME AG
> Talacker 50
> 8001 Zurich, Switzerland
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> For additional commands, e-mail: users-help@artemis.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
For additional commands, e-mail: users-help@artemis.apache.org
Re: Slow connects after updating to 2.57.0
Posted by Thorsten Meinl <th...@knime.com>.
Am Dienstag, dem 22.09.2026 um 08:43 -0500 schrieb Justin Bertram:
> Did you move the broker from 2.56.0 to 2.57.0?
We updated from 2.55.0 to 2.57.0.
Attached are two logs from an affected service. The only difference is
the broker version it connects to (2.55 vs 2.57). As you can see, with
2.57 is takes considerably longer between the two INFO message (~1s vs.
~9s). In the 2.57 case I see remote destinations such as
localhost/127.0.0.1:61616 or /0.0.0.0:61616 being tried. Which
obviously doesn't work and seems to me to be the cause for the delays.
Maybe this information helps.
Thanks,
Thorsten
> On Tue, Sep 22, 2026 at 5:52 AM Thorsten Meinl
> <th...@knime.com> wrote:
> >
> > Hello everyone,
> >
> > After updating our Artemis broker to 2.57.0 we are experiencing
> > high
> > latencies when clients connect to the broker via the CORE protocol.
> > The
> > clients themselves have not been changed. They are currently using
> > version 2.52.0 of the JMS client lib (pulled in by a dependency
> > chain).
> > Client are all authenticated and connect via
> > jakarta.jms.ConnectionFactory.createContext().
> >
> > After reading the release notes, we believe it has something to do
> > with
> > the changed Core protocol connect handshake. However, it is also
> > mentioned that the change is "backward-compatible" and therefore we
> > did
> > not pay much attention. So a couple of questions:
> >
> > - Could this broker change be responsible for the increased
> > connection
> > times?
> > - If so, how can we fix it without changing all clients?
> > - Is this effect expected/intended?
> >
> > Please let me know, if you need more information.
> >
> > Thanks and regards,
> >
> > Thorsten
> >
> > --
> > Dr.-Ing. Thorsten Meinl
> > KNIME AG
> > Talacker 50
> > 8001 Zurich, Switzerland
> >
> >
> > -------------------------------------------------------------------
> > --
> > To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> > For additional commands, e-mail: users-help@artemis.apache.org
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> For additional commands, e-mail: users-help@artemis.apache.org
--
Dr.-Ing. Thorsten Meinl
KNIME AG
Talacker 50
8001 Zurich, Switzerland
Re: Slow connects after updating to 2.57.0
Posted by Justin Bertram <jb...@apache.org>.
Did you move the broker from 2.56.0 to 2.57.0?
> Could this broker change be responsible for the increased connection times?
At this point I don't see how it could be responsible. If your clients
didn't change then the handshake also didn't change in any meaningful
way.
> Is this effect expected/intended?
Clients using the new Core connection handshake will technically see
slightly longer connection times due to the new authentication path,
but overall there shouldn't be a meaningful increase because this work
was already being done just later in the connection process.
Justin
On Tue, Sep 22, 2026 at 5:52 AM Thorsten Meinl <th...@knime.com> wrote:
>
> Hello everyone,
>
> After updating our Artemis broker to 2.57.0 we are experiencing high
> latencies when clients connect to the broker via the CORE protocol. The
> clients themselves have not been changed. They are currently using
> version 2.52.0 of the JMS client lib (pulled in by a dependency chain).
> Client are all authenticated and connect via
> jakarta.jms.ConnectionFactory.createContext().
>
> After reading the release notes, we believe it has something to do with
> the changed Core protocol connect handshake. However, it is also
> mentioned that the change is "backward-compatible" and therefore we did
> not pay much attention. So a couple of questions:
>
> - Could this broker change be responsible for the increased connection
> times?
> - If so, how can we fix it without changing all clients?
> - Is this effect expected/intended?
>
> Please let me know, if you need more information.
>
> Thanks and regards,
>
> Thorsten
>
> --
> Dr.-Ing. Thorsten Meinl
> KNIME AG
> Talacker 50
> 8001 Zurich, Switzerland
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> For additional commands, e-mail: users-help@artemis.apache.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
For additional commands, e-mail: users-help@artemis.apache.org
Re: Slow connects after updating to 2.57.0
Posted by Alexander Milovidov <mi...@gmail.com>.
I have configured an external availability test which is quite simple: it
connects to Artemis, sends/receives 100 messages and disconnects. It
performs once in a minute and exports some metrics, for example -
connection time, produce/consume time. We have several primary/backup
clusters and test availability of each cluster.
There were no significant changes after upgrading these clusters to 2.57.0.
The connection time remain the same. The client used in the availability
test utility is quite old (2.41.0), I have to build a new version.
вт, 22 сент. 2026 г., 13:52 Thorsten Meinl <th...@knime.com>:
> Hello everyone,
>
> After updating our Artemis broker to 2.57.0 we are experiencing high
> latencies when clients connect to the broker via the CORE protocol. The
> clients themselves have not been changed. They are currently using
> version 2.52.0 of the JMS client lib (pulled in by a dependency chain).
> Client are all authenticated and connect via
> jakarta.jms.ConnectionFactory.createContext().
>
> After reading the release notes, we believe it has something to do with
> the changed Core protocol connect handshake. However, it is also
> mentioned that the change is "backward-compatible" and therefore we did
> not pay much attention. So a couple of questions:
>
> - Could this broker change be responsible for the increased connection
> times?
> - If so, how can we fix it without changing all clients?
> - Is this effect expected/intended?
>
> Please let me know, if you need more information.
>
> Thanks and regards,
>
> Thorsten
>
> --
> Dr.-Ing. Thorsten Meinl
> KNIME AG
> Talacker 50
> 8001 Zurich, Switzerland
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@artemis.apache.org
> For additional commands, e-mail: users-help@artemis.apache.org
>
>
>