zookeeper-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jerry Hebert <jerry.heb...@gmail.com>
Subject "Connections" incorrectly reported as 1 (3.5.5)
Date Thu, 24 Oct 2019 22:52:01 GMT
Hey all,

I've upgraded one of my ensembles to 3.5.5 now
(3.5.5-390fe37ea45dee01bf87dc1c042b5e3dcce88653, specifically). All of my
metrics appeared to be healthy but after the migration, I noticed that the
new ensemble has *all 5 nodes reporting a connection count of 1* (via the
"stat" 4ltr command as well as zk_num_alive_connections from the "mntr"

The servers are clearly receiving traffic: I can see node counts going up
and down and I can see clients making changes to various keys. I can also
monitor netstat for 2181 connections and again see connections fluctuating
per usual but I still see "Connections: 1" in stats. This translates into
our Datadog agent reporting connections as 1 too. I've been reading through
the code to try to understand how this may be possible but it's a bit of a
slog as I'm unfamiliar with it and I've found myself digging into Netty now.

I pasted a couple of possibly relevant log lines below. In particular, note
that "secure" is false here and I noticed that the conx count is split in
the code depending on whether or not you're in secure mode. I also find it
odd that I'm seeing 0:0:0:0:0:0:0:0 in the logs which looks like ipv6 to me
and I'm using ipv4 (or at least, I partially am...). I also don't
understand the zxid expectation mismatch.

2019-10-24 22:12:08,411 [myid:9] - WARN
- Got zxid 0x2f00000001 expected 0x1

2019-10-24 22:11:57,349 [myid:9] - INFO  [main:ServerCnxnFactory@135] -
Using org.apache.zookeeper.server.NIOServerCnxnFactory as server connection

Any advice would be greatly appreciated. I don't feel comfortable leaving
this server as-is given that it's misreporting connections. Something is
definitely wrong.

Thanks in advance!


  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message