Description
On a previous investigation of a BN in Previewnet trying to backfill, there were some connections that timed-out, and also lots of blocks that were received unabled to be parsed, with different parsing issues.
When manually querying the source and the block ranges, I was able to get the blocks, some times after a single retry but most of the time at first. This manual test was done using Postman (not helidon) webClient.
Steps to Reproduce
There is still not a reliable way to reproduce this, the current hypothesis is that HTTP2 WebClient of Helidon is not robust when network is choppy or has fluctuations, resulting in half-dead connections (we have seen this with the publisher as a form of stalled connections) and parsing issues.
The first part of this issue is to be able to find the root cause by introducing network fluctuations and chaos.
One idea is to use Chaos Mesh to introduce some network conditions and see if we can reliably reproduce the error. finally create an isolated test that is able to identify the issue, and if possible to fix it.
Additional context
These are some of the error messages found:
Failed to fetch blocks [151,290->151,309] from node [[lfh02.previewnet.blocknode.hashgraph-devops.com](http://lfh02.previewnet.blocknode.hashgraph-devops.com/)] (attempt 1/3) due to java.lang.RuntimeException: Error fetching blocks - , com.hedera.pbj.runtime.ParseException: java.io.IOException: Cannot understand wire_type of 6.
Failed to fetch blocks [0->19] from node [[lfh02.previewnet.blocknode.hashgraph-devops.com](http://lfh02.previewnet.blocknode.hashgraph-devops.com/)] (attempt 3/3) due to java.lang.RuntimeException: Error fetching blocks - , java.lang.RuntimeException: Timed out waiting for block stream response.
warn rfh01 2026-04-17 20:19:18.250+0000 WARNING [io.helidon.webclient.http2.Http2ClientConnection handle] Unsupported frame type!! UNKNOWN
[org.hiero.block.node.backfill.BackfillFetcher fetchBlocksFromNode] Failed to fetch blocks [0->19] from node [[lfh02.previewnet.blocknode.hashgraph-devops.com](http://lfh02.previewnet.blocknode.hashgraph-devops.com/)] (attempt 1/3) due to java.lang.RuntimeException: Error fetching blocks - , com.hedera.pbj.runtime.ParseException: java.io.IOException: Bad tag [8], field [1] wireType [0].
Description
On a previous investigation of a BN in Previewnet trying to backfill, there were some connections that timed-out, and also lots of blocks that were received unabled to be parsed, with different parsing issues.
When manually querying the source and the block ranges, I was able to get the blocks, some times after a single retry but most of the time at first. This manual test was done using Postman (not helidon) webClient.
Steps to Reproduce
There is still not a reliable way to reproduce this, the current hypothesis is that HTTP2 WebClient of Helidon is not robust when network is choppy or has fluctuations, resulting in half-dead connections (we have seen this with the publisher as a form of stalled connections) and parsing issues.
The first part of this issue is to be able to find the root cause by introducing network fluctuations and chaos.
One idea is to use Chaos Mesh to introduce some network conditions and see if we can reliably reproduce the error. finally create an isolated test that is able to identify the issue, and if possible to fix it.
Additional context
These are some of the error messages found: