Test installation on aarch64 - Failed to report error: Error connecting to Redis on localhost:6379

Product Support Self-hosting

AI-generated summary

The discussion revolves around resolving the error “Failed to report error: Error connecting to Redis on localhost:6379” encountered during the test installation of Discourse on aarch64 architecture.

Initially, Falco suspects another process is using the port, but esk8builders shares the output of lsof -i -P -n | grep LISTEN which shows no other process is listening on port 6379.

Falco then discovers that the issue is due to Redis not starting up, and attributes this to the system page size not being supported by jemalloc. esk8builders shares the system page size as 65536, which is not compatible with the 4096 page size used by jemalloc.

Falco suggests that the issue might be specific to the Oracle Linux kernel and recommends using a Debian-based OS. xwu reports the same issue on Oracle’s Ampere server and pfaffman advises trying a Debian-based OS as a potential solution.

1

Hello All,

I am trying to test installation on aarch64 - Linux XXXXXX 5.4.17-2136.302.6.1.el8uek.aarch64 #2 SMP Tue Dec 14 12:00:21 PST 2021 aarch64 aarch64 aarch64 GNU/Linux

and following errors appear:

I, [2021-12-21T11:04:26.588172 #1] INFO – : > cd /var/www/discourse && su discourse -c ‘bundle exec rake plugin:pull_compatible_all’
I, [2021-12-21T11:04:29.293636 #1] INFO – : docker_manager is already at latest compatible version
discourse-chat-integration is already at latest compatible version

I, [2021-12-21T11:04:29.293942 #1] INFO – : > cd /var/www/discourse && su discourse -c ‘bundle exec rake db:migrate’
Failed to report error: Error connecting to Redis on localhost:6379 (Errno::EADDRNOTAVAIL) 2 Error connecting to Redis on localhost:6379 (Errno::EADDRNOTAVAIL) subscribe failed, reconnecting in 1 second. Call stack /var/www/discourse/vendor/bundle/ruby/2.7.0/gems/redis-4.5.1/lib/redis/client.rb:398:in rescue in establish_connection' /var/www/discourse/vendor/bundle/ruby/2.7.0/gems/redis-4.5.1/lib/redis/client.rb:379:in establish_connection’

Please, can you share any tips on how to resolve this error?
Thanks a lot!

2

That means that there is something already in that port. Can you stop whatever is listening there?

3

Thanks for your prompt reply!

I did some checks but its ok, now I am attaching full log file from the setup

discourse_errors.txt (73.7 KB)

Thanks a lot!

4

Can you run sudo lsof -i -P -n | grep LISTEN on your host and share the output here?

5

]# lsof -i -P -n | grep LISTEN
systemd 1 root 36u IPv4 19779 0t0 TCP *:111 (LISTEN)
systemd 1 root 38u IPv6 19781 0t0 TCP *:111 (LISTEN)
rpcbind 1507 rpc 4u IPv4 19779 0t0 TCP *:111 (LISTEN)
rpcbind 1507 rpc 6u IPv6 19781 0t0 TCP *:111 (LISTEN)
sshd 2159 root 5u IPv4 33431 0t0 TCP *:22 (LISTEN)
sshd 2159 root 6u IPv6 33441 0t0 TCP *:22 (LISTEN)
osms-agen 2488 root 6u IPv6 39741 0t0 TCP 127.0.0.1:32768 (LISTEN)
osms-agen 2488 root 13u IPv4 40406 0t0 TCP 127.0.0.1:9003 (LISTEN)

6

Looks like I was wrong here. The problem is that redis isn’t starting up…

<jemalloc>: Unsupported system page size
<jemalloc>: Unsupported system page size

@esk8builders can you please run

uname -a
getconf PAGE_SIZE
7

Linux 5.4.17-2136.302.6.1.el8uek.aarch64 #2 SMP Tue Dec 14 12:00:21 PST 2021 aarch64 aarch64 aarch64 GNU/Linux

65536

8

What distro is this? I tested the aarch64 using the official Raspbian distro.

We use jemalloc with 4096 page size, that is why it fails on your configured page size.

9

4096
same here, on a standard ubuntu 20.04 on Oracle cloud arm server

10

Hello Falco, This is standard for [Oracle-Linux-8.5-aarch64-2021.12.14-0] for Ampere server processors

I will try to change to more standard architecture

Thanks for your help!

11

Hmm so the Oracle Linux kernel was compiled with this somewhat different PAGE_SIZE.

Can you run a debian distro in those Ampere servers @esk8builders ?

12

This is likely due to the CPU hardware - smaller page sizes aren’t available or are extremely inefficient at maximum RAM

13

Hi, I’m having the same issue when trying to install discourse via docker on Oracle’s Ampere server, spec is
Oracle Linux Server 8.6
Linux 5.4.17-2136.305.5.4.el8uek.aarch64

Is there a solution now? Any suggestion on what I should try? Thanks a lot!

14

Did you try using a Debian based os as recommended above?