I'm confused about something. Can any postfix experts help debug?
On Friday I put a new package-owner-alias cronjob in place on bastion (that Matt Prahl wrote). It generates the package owner list from pagure over dist-git instead of generating it from pkgdb.
https://infrastructure.fedoraproject.org/cgit/ansible.git/commit/?id=eb241f9...
Kevin saw that the hourly cronjob is spitting out some errors from the postfix command.
However, when I inspect the file, I don't find any duplicates:
$ cat /etc/postfix/package-owner | awk ' { print $1 } ' | uniq -d
Anybody know what's wrong with that file/output?
-Ralph
I'm confused about something. Can any postfix experts help debug?
On Friday I put a new package-owner-alias cronjob in place on bastion (that Matt Prahl wrote). It generates the package owner list from pagure over dist-git instead of generating it from pkgdb.
https://infrastructure.fedoraproject.org/cgit/ansible.git/commit/?id=eb24...
Kevin saw that the hourly cronjob is spitting out some errors from the postfix command.
However, when I inspect the file, I don't find any duplicates:
$ cat /etc/postfix/package-owner | awk ' { print $1 } ' | uniq -dAnybody know what's wrong with that file/output?
The problem is not in the file/output. It's in the running of the script. If Pagure on Dist-Git takes too long to reply, or is down, haproxy returns an HTML page with "this app is offline".
That is what's causing the failure: simplejson.scanner.JSONDecodeError: Expecting value: line 2 column 1 (char 1) error creating owner-alias file
Given that the script asks Pagure to page through a lot of projects everytime, the chance of any one of those requests failing gets quite high. Thus, basically every run crashes some part through the run because a Pagure result took too long or was aborted by haproxy.
-Ralph
On Wed, Aug 09, 2017 at 05:24:27PM -0000, Patrick マルタインアンドレアス Uiterwijk wrote:
I'm confused about something. Can any postfix experts help debug?
On Friday I put a new package-owner-alias cronjob in place on bastion (that Matt Prahl wrote). It generates the package owner list from pagure over dist-git instead of generating it from pkgdb.
https://infrastructure.fedoraproject.org/cgit/ansible.git/commit/?id=eb24...
Kevin saw that the hourly cronjob is spitting out some errors from the postfix command.
However, when I inspect the file, I don't find any duplicates:
$ cat /etc/postfix/package-owner | awk ' { print $1 } ' | uniq -dAnybody know what's wrong with that file/output?
The problem is not in the file/output. It's in the running of the script. If Pagure on Dist-Git takes too long to reply, or is down, haproxy returns an HTML page with "this app is offline".
That is what's causing the failure: simplejson.scanner.JSONDecodeError: Expecting value: line 2 column 1 (char 1) error creating owner-alias file
Given that the script asks Pagure to page through a lot of projects everytime, the chance of any one of those requests failing gets quite high. Thus, basically every run crashes some part through the run because a Pagure result took too long or was aborted by haproxy.
We're speaking about two different issues here: 1/ pagure returning 503 a certain number of time in a row making the entire script crash 2/ these errors: ``` /etc/cron.hourly/package-owner-aliases.sh:
postalias: warning: /etc/postfix/package-owner.db: duplicate entry: "perl-mogilefs-server-owner" postalias: warning: /etc/postfix/package-owner.db: duplicate entry: "simplyhtml-owner" postalias: warning: /etc/postfix/package-owner.db: duplicate entry: "packagekit-qt-owner" postalias: warning: /etc/postfix/package-owner.db: duplicate entry: "php-pimple-owner" postalias: warning: /etc/postfix/package-owner.db: duplicate entry: "vlgothic-fonts-owner" ```
I believe these are the ones Ralph was referring to, not the former.
Pierre
On Wed, Aug 9, 2017 at 7:29 PM, Pierre-Yves Chibon pingou@pingoured.fr wrote:
On Wed, Aug 09, 2017 at 05:24:27PM -0000, Patrick マルタインアンドレアス Uiterwijk wrote:
I'm confused about something. Can any postfix experts help debug?
On Friday I put a new package-owner-alias cronjob in place on bastion (that Matt Prahl wrote). It generates the package owner list from pagure over dist-git instead of generating it from pkgdb.
https://infrastructure.fedoraproject.org/cgit/ansible.git/commit/?id=eb24...
Kevin saw that the hourly cronjob is spitting out some errors from the postfix command.
However, when I inspect the file, I don't find any duplicates:
$ cat /etc/postfix/package-owner | awk ' { print $1 } ' | uniq -dAnybody know what's wrong with that file/output?
The problem is not in the file/output. It's in the running of the script. If Pagure on Dist-Git takes too long to reply, or is down, haproxy returns an HTML page with "this app is offline".
That is what's causing the failure: simplejson.scanner.JSONDecodeError: Expecting value: line 2 column 1 (char 1) error creating owner-alias file
Given that the script asks Pagure to page through a lot of projects everytime, the chance of any one of those requests failing gets quite high. Thus, basically every run crashes some part through the run because a Pagure result took too long or was aborted by haproxy.
We're speaking about two different issues here: 1/ pagure returning 503 a certain number of time in a row making the entire script crash 2/ these errors:
/etc/cron.hourly/package-owner-aliases.sh: postalias: warning: /etc/postfix/package-owner.db: duplicate entry: "perl-mogilefs-server-owner" postalias: warning: /etc/postfix/package-owner.db: duplicate entry: "simplyhtml-owner" postalias: warning: /etc/postfix/package-owner.db: duplicate entry: "packagekit-qt-owner" postalias: warning: /etc/postfix/package-owner.db: duplicate entry: "php-pimple-owner" postalias: warning: /etc/postfix/package-owner.db: duplicate entry: "vlgothic-fonts-owner"I believe these are the ones Ralph was referring to, not the former.
Oh. That's because they're in /etc/aliases, generated by fasClient.
Pierre _______________________________________________ infrastructure mailing list -- infrastructure@lists.fedoraproject.org To unsubscribe send an email to infrastructure-leave@lists.fedoraproject.org
On Wed, Aug 9, 2017 at 7:32 PM, Patrick Uiterwijk puiterwijk@redhat.com wrote:
On Wed, Aug 9, 2017 at 7:29 PM, Pierre-Yves Chibon pingou@pingoured.fr wrote:
On Wed, Aug 09, 2017 at 05:24:27PM -0000, Patrick マルタインアンドレアス Uiterwijk wrote:
I'm confused about something. Can any postfix experts help debug?
On Friday I put a new package-owner-alias cronjob in place on bastion (that Matt Prahl wrote). It generates the package owner list from pagure over dist-git instead of generating it from pkgdb.
https://infrastructure.fedoraproject.org/cgit/ansible.git/commit/?id=eb24...
Kevin saw that the hourly cronjob is spitting out some errors from the postfix command.
However, when I inspect the file, I don't find any duplicates:
$ cat /etc/postfix/package-owner | awk ' { print $1 } ' | uniq -dAnybody know what's wrong with that file/output?
The problem is not in the file/output. It's in the running of the script. If Pagure on Dist-Git takes too long to reply, or is down, haproxy returns an HTML page with "this app is offline".
That is what's causing the failure: simplejson.scanner.JSONDecodeError: Expecting value: line 2 column 1 (char 1) error creating owner-alias file
Given that the script asks Pagure to page through a lot of projects everytime, the chance of any one of those requests failing gets quite high. Thus, basically every run crashes some part through the run because a Pagure result took too long or was aborted by haproxy.
We're speaking about two different issues here: 1/ pagure returning 503 a certain number of time in a row making the entire script crash 2/ these errors:
/etc/cron.hourly/package-owner-aliases.sh: postalias: warning: /etc/postfix/package-owner.db: duplicate entry: "perl-mogilefs-server-owner" postalias: warning: /etc/postfix/package-owner.db: duplicate entry: "simplyhtml-owner" postalias: warning: /etc/postfix/package-owner.db: duplicate entry: "packagekit-qt-owner" postalias: warning: /etc/postfix/package-owner.db: duplicate entry: "php-pimple-owner" postalias: warning: /etc/postfix/package-owner.db: duplicate entry: "vlgothic-fonts-owner"I believe these are the ones Ralph was referring to, not the former.
Oh. That's because they're in /etc/aliases, generated by fasClient.
Wrong again on my part. It was case sensitivity: for postfix, MySQL-owner and mysql-owner are the same alias, since email addresses are case insensitive.
Pierre _______________________________________________ infrastructure mailing list -- infrastructure@lists.fedoraproject.org To unsubscribe send an email to infrastructure-leave@lists.fedoraproject.org
Introducing some retry stuff to be more resilient when pagure blips: https://infrastructure.fedoraproject.org/cgit/ansible.git/commit/?id=b084a77...
infrastructure@lists.fedoraproject.org
-
Patrick マルタインアンドレアス Uiterwijk -
Patrick Uiterwijk
-
Pierre-Yves Chibon -
Ralph Bean