This is pretty much not a bug in Matomo, so the forums would be a better place for this.

Anyway, please take a look into your PHP error log from Matomo to see if there is anything related to the triggered archiving requests.

Ok, thanks. Added a new forum topic then:
https://forum.matomo.org/t/gateway-timout-issue-on-cron-archive-task/32583

Unfortunately, nothing shows up in the error logfile related to this.

@fdellwing

How do we know that this is not a bug in Matomo?

Its simple: If this would be a bug with Matomo, much more users would have it. Furthermore, a gateway timeout is a typical problem for a misconfigured PHP instance.

I am having a similar issue.

The warning message is this:

WARNING [2019-05-02 12:38:48] 2017 /opt/matomo/core/Common.php(271): Notice - unserialize(): Error at offset 0 of 91 bytes - Matomo 3.9.1 - Please report this message in the Matomo forums: https://forum.matomo.org (please do a search first as it might have been reported already)

As you can see, the warning message has this unserialize() that appeared after Matomo 3.9.0. Might have nothing to do with this, I just mention it.

After the warning message, the error message is this:

ERROR [2019-05-02 12:38:52] 2017 Empty or invalid response

504 Gateway Time-out

The server didn't respond in time. for website id 19, Time elapsed: 39.120s, skipping

The thing is that this problem is not consistent with all the websites and all the periods. Right now I am trying to reprocess all the logs and this error doesn't appear after every debug but occasionally.

If the problem is a misconfigured PHP instance, then where should I look for the possible causes/solutions? I have just an installation of Software Collections PHP 7.2 by RedHat and the only values altered in php.ini are the memory_limit and max_execution_time which I have increased.

The unserialize error will always be shown if the answer from the server is not the expected, so the problem is unrelated.

What webserver are you using? Is there a reverse proxy in place? Do you use PHP-fpm (probably)?

I am using Apache httpd deployed in Openshift. There is a reverse proxy, not configured by me. I don't know if I am using PHP-fpm, I didn't make any changes in PHP besides installing PHP and changing the two variables mentioned.

The installed packages are:

rh-php72-php \
rh-php72-php-curl \
rh-php72-php-gd \
rh-php72-php-cli \
rh-php72-php-mysqlnd \
rh-php72-php-mbstring \
rh-php72-php-xml

Ok, that is a setup I cannot directly help you with. But I can say you one thing: Most webservers have a default timeout configured for proxy handling. You can override this setting, for apache see https://httpd.apache.org/docs/2.4/mod/mod_proxy.html#proxytimeout.

I prefer a timeout one second shorter than your PHP max_execution_time.

For nginx with php-fpm this would look like fastcgi_read_timeout 119 if the max_execution_time ist 120 seconds.

I am not using fpm thought. Could it be the reason I am seeing this error?

Could it be the reason I am seeing this error?

Unlikely.

There is a reverse proxy, not configured by me.

All I said is valid for all proxies, php-fpm is just a very good example.

Sometimes I am also getting:

Oops… there was a problem during the request. Maybe the server had a temporary issue, or maybe you requested a report with too much data. Please try again. If this error occurs repeatedly please contact your Matomo administrator for assistance.

In widgets (eg Visits Over Time) if I request reports for a whole year. The websites are not archived because archiving fails in many sites with the discussed error.

Maybe these issues are related but I have no idea what might be wrong in the server configuration. What do you think?

Also, I should mention that the database without any archived table is around 200GB.