[
https://issues.apache.org/jira/browse/RATIS-2185?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17895038#comment-17895038
]
Tsz-wo Sze edited comment on RATIS-2185 at 11/2/24 5:39 PM:
------------------------------------------------------------
{quote}... some vague message like "Unavailable" or "Client cancelled". ...
{quote}
These messages are provided by the StatusRuntimeException from gRPC. Ratis may
be able to somehow detect such cases and then print some better messages.
{quote}... We've seen examples where "Client cancelled" is displayed for
technical problem in GRPC layer, e.g. when appendEntries buffers are bigger
than the max GRPC acceptable message size.
{quote}
The message is the following StatusRuntimeException from gRPC as described in
RATIS-2135.
- StatusRuntimeException: CANCELLED: RST_STREAM closed stream. HTTP/2 error
code: CANCEL
It is unfortunate that gRPC throws such an unclear message. We probably should
file a bug for the gRPC community.
was (Author: szetszwo):
bq. ... some vague message like "Unavailable" or "Client cancelled". ...
These messages are provided by the StatusRuntimeException from gRPC. Ratis may
be able to somehow detect such cases and then print some better messages.
bq. ... We've seen examples where "Client cancelled" is displayed for technical
problem in GRPC layer, e.g. when appendEntries buffers are bigger than the max
GRPC acceptable message size.
The message is the following StatusRuntimeException from gRPC as described in
RATIS-2135.
- StatusRuntimeException: CANCELLED: RST_STREAM closed stream. HTTP/2 error
code: CANCEL
It is unfortunate that gRPC throws such an unclear message.
> Ratis logs debugability
> -----------------------
>
> Key: RATIS-2185
> URL: https://issues.apache.org/jira/browse/RATIS-2185
> Project: Ratis
> Issue Type: Improvement
> Reporter: Duong
> Priority: Major
>
> Created this as a umbrella ticket for all changes related to improve the
> debugability of Ratis logs.
> Today, there're a couple of issues with ratis logs making debug issues in
> ratis a hard job. For example:
> * Lack of details: when an RPC call, e.g. appendEntries calls, fail, there
> is very little information to identify the problem, exception some vague
> message like "Unavailable" or "Client cancelled". We've seen examples where
> "Client cancelled" is displayed for technical problem in GRPC layer, e.g.
> when appendEntries buffers are bigger than the max GRPC acceptable message
> size.
> * Repetitive: TBF
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
