-
Notifications
You must be signed in to change notification settings - Fork 138
[BUG] KoP failed to start for PulsarAdminException #619
Comments
KOP:2.7.0.1 |
出问题的时候,每个broker节点都报类似的错误,__consumer_offsets 操作30秒超时,后面把所有的broker停掉之后,只启动一台,启动前,注释掉broker.conf里面的kop配置,然后再启动,broker恢复正常。 另外,重启过程中话发下 /tmp/pulsar-nar 下面,少了部分类,删掉后,重启恢复 When a problem happened, each broker node reported a similar error. The __consumer_offsets operation timed out for 30 seconds. After stopping all the brokers, only start one broker. Before starting, comment out the kop configuration in broker.conf, and then the broker started successfully. In addition, during the restarting process, send a message under /tmp/pulsar-nar. Some classes are missing. After deleting, restart and restore |
This bug is caused by kop/kafka-impl/src/main/java/io/streamnative/pulsar/handlers/kop/utils/MetadataUtils.java Line 155 in 96f454a
However, the implementation of broker side is to create an admin client and send HTTP request to itself. There's somehow a deadlock for this implementation. Here're a correct output for
The The simplest fix is to remove the |
) Fixes #619 See #619 (comment) for details.
It's fixed by #626 |
The text was updated successfully, but these errors were encountered: