github-actions[bot] commented on code in PR #70693:
URL: https://github.com/apache/doris/pull/70693#discussion_r4637801570
##########
be/src/exec/sink/writer/vtablet_writer.cpp:
##########
@@ -604,8 +604,12 @@ Status VNodeChannel::init(RuntimeState* state) {
_row_desc = std::make_unique<RowDescriptor>(_tuple_desc);
_batch_size = state->batch_size();
- _stub =
state->exec_env()->brpc_internal_client_cache()->get_client(_node_info.host,
-
_node_info.brpc_port);
+ // Remote catalog writes retain the remote cluster's transport policy,
independently of
+ // whether private RPC TLS is enabled in this cluster.
+ _stub = state->exec_env()->brpc_internal_client_cache()->get_client(
+ _node_info.host, _node_info.brpc_port,
+ _parent->_t_sink.olap_table_sink.remote_endpoint ?
client::RpcPurpose::EXTERNAL_SERVICE
Review Comment:
[P1] Preserve remote-write transport policy for old FE payloads.
`remote_endpoint` is optional and is absent when a pre-change FE serializes
`RemoteOlapTableSink`, but that payload still carries
`partition.master_address` (the legacy remote marker). The changed
VNodeChannel, HTTP-batch, streaming, and row-distribution paths test only this
new bool, so a TLS-enabled new BE treats an old-FE remote insert as
`INTERNAL_RPC` and sends TLS to the plaintext remote BE/FE, failing the write
before the remote cluster sees it. Derive the external policy from the legacy
marker whenever the new field is absent and cover an old-style sink in a
rolling-upgrade test.
##########
fe/be-java-extensions/max-compute-connector/src/main/java/org/apache/doris/maxcompute/MaxComputeFeClient.java:
##########
@@ -203,12 +211,6 @@ private void sleepBeforeRetry() throws IOException {
}
}
- private static TTransport createTransport(TNetworkAddress address, int
timeoutMs,
Review Comment:
[P1] Accept the native PKCS#1 key format in the MaxCompute TLS client. The
BE forwards `tls_private_key_path` unchanged, and this call passes it directly
to Netty's file-based `keyManager`, which only parses PKCS#8 PEM. The PR's
BouncyCastle dependency is not used to convert or parse the key, so a normal
`BEGIN RSA PRIVATE KEY` identity makes `createTransport` fail and every
block-id allocation aborts under private TLS; the added PKCS#1 regression
currently exercises that same failing path. Parse/convert PKCS#1 before
building the Netty context (or otherwise use a `PrivateKey`) and retain the
negative tests.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]