[
https://issues.apache.org/jira/browse/IGNITE-27554?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Aleksandr Chesnokov updated IGNITE-27554:
-----------------------------------------
Description:
h2. Problem
When a node with a different Ignite version tries to join (e.g. 2.17 to 2.18),
the discovery handshake may fail before node validation.
We expect a clear message like:
"Remote node rejected due to incompatible version for cluster join"
but it is not printed. Instead, logs show low-level errors like
StreamCorruptedException, EOFException, or Connection reset, so the real reason
(version mismatch) is unclear.
The root cause is that the discovery message serialization format changed:
after IGNITE_HEADER the new protocol writes an extra leading byte (serMode),
while older nodes do not write it. As a result, the new node cannot even read
TcpDiscoveryHandshakeRequest and logs:
"Received unexpected byte while reading discovery message: -84"
This message is confusing for support and does not explain what action to take.
h2. Steps to reproduce
# Build latest version of Ignite. Let's call it new node
# Build Ignite from commit 354ac793. Let's call it old node
# Try to connect them
was:
h2. Problem
When a node with a different Ignite version tries to join (e.g. 2.17 to 2.18),
the discovery handshake may fail before node validation.
We expect a clear message like:
"Remote node rejected due to incompatible version for cluster join"
but it is not printed. Instead, logs show low-level errors like
StreamCorruptedException, EOFException, or Connection reset, so the real reason
(version mismatch) is unclear.
The root cause is that the discovery message serialization format changed:
after IGNITE_HEADER the new protocol writes an extra leading byte (serMode),
while older nodes do not write it. As a result, the new node cannot even read
TcpDiscoveryHandshakeRequest and logs:
"Received unexpected byte while reading discovery message: -84"
This message is confusing for support and does not explain what action to take.
> Improve discovery handshake diagnostics for incompatible versions
> -----------------------------------------------------------------
>
> Key: IGNITE-27554
> URL: https://issues.apache.org/jira/browse/IGNITE-27554
> Project: Ignite
> Issue Type: Improvement
> Reporter: Aleksandr Chesnokov
> Assignee: Aleksandr Chesnokov
> Priority: Major
>
> h2. Problem
> When a node with a different Ignite version tries to join (e.g. 2.17 to
> 2.18), the discovery handshake may fail before node validation.
> We expect a clear message like:
> "Remote node rejected due to incompatible version for cluster join"
> but it is not printed. Instead, logs show low-level errors like
> StreamCorruptedException, EOFException, or Connection reset, so the real
> reason (version mismatch) is unclear.
> The root cause is that the discovery message serialization format changed:
> after IGNITE_HEADER the new protocol writes an extra leading byte (serMode),
> while older nodes do not write it. As a result, the new node cannot even read
> TcpDiscoveryHandshakeRequest and logs:
> "Received unexpected byte while reading discovery message: -84"
> This message is confusing for support and does not explain what action to
> take.
> h2. Steps to reproduce
> # Build latest version of Ignite. Let's call it new node
> # Build Ignite from commit 354ac793. Let's call it old node
> # Try to connect them
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
