Wednesday, June 10, 2020

Retry wagtail as a CMS

I think we are going to use a wagtail site for the laudate project. In theory I would prefer to do it with a Lino site, but reality sometimes differs from theory: the users of the laudate project won’t have time to actively contribute to development. And why not have a look over the fence into other projects.

https://docs.wagtail.io/en/stable/getting_started

First steps on the laudate server:

$ cd /usr/local/lino/lino_local
$ mkdir laudate
$ cd laudate
$ virtualenv -p python3 env
$ a
$ pip install wagtail
$ wagtail start mysite
Creating a Wagtail project called mysite
Success! mysite has been created

$ cd mysite
$ pip install -r requirements.txt
$ ./manage.py migrate
$ ./manage.py createsuperuser
$ ./manage.py runserver

I noticed that it requires Django<3.1

I then noticed that this quick start is for a developer site. Need to adapt the settings for a production site.

How to properly configure postfix and mailman

I continued (partly with Tonis and partly with Hamza) to try to get the mail systems running properly. I started a new section Setting up a mail server in the hoster’s guide of the book.

Now we get another error from the recipient’s SMTP server:

status=deferred (host mx01.emig.gmx.net[212.227.17.5] refused to talk to me: 554-gmx.net (mxgmx115) Nemesis ESMTP Service not available 554-No SMTP service 554-Bad DNS PTR resource record. 554 For explanation visit https://www.gmx.net/mail/senderguidelines?ip=95.217.218.29&c=rdns)

How to test an SMTP server:

$ telnet lino-framework.org 465
Trying 167.114.229.225...
Connected to lino-framework.org.
Escape character is '^]'.

For laudate we had to uncomment the smtps entry in /etc/postfix/master.cf:

smtps     inet  n       -       y       -       -       smtpd

Mxtools said “Reverse DNS is not a Valid Hostname” for laudate. That’s fixed now.

TIL that also reverse DNS has a propagation time.

TIL that I can say dig to do a DNS lookup, and dig -x to do a reverse DNS lookup. dig laudate.ee gives me 95.217.218.29 and dig -x 95.217.218.29 gives me laudate.ee:

$ dig laudate.ee
; <<>> DiG 9.11.3-1ubuntu1.12-Ubuntu <<>> laudate.ee
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 58844
;; flags: qr rd ra; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 1

;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags:; udp: 65494
;; QUESTION SECTION:
;laudate.ee.                  IN      A

;; ANSWER SECTION:
laudate.ee.           3600    IN      A       95.217.218.29

;; Query time: 12 msec
;; SERVER: 127.0.0.53#53(127.0.0.53)
;; WHEN: Wed Jun 10 17:03:36 EEST 2020
;; MSG SIZE  rcvd: 55

$ dig -x 95.217.218.29

; <<>> DiG 9.11.3-1ubuntu1.12-Ubuntu <<>> -x 95.217.218.29
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 26604
;; flags: qr rd ra; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 1

;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags:; udp: 65494
;; QUESTION SECTION:
;29.218.217.95.in-addr.arpa.  IN      PTR

;; ANSWER SECTION:
29.218.217.95.in-addr.arpa. 72562 IN  PTR     mail.laudate.ee.

;; Query time: 6 msec
;; SERVER: 127.0.0.53#53(127.0.0.53)
;; WHEN: Wed Jun 10 17:06:59 EEST 2020
;; MSG SIZE  rcvd: 84

But why does it say mail.laudate.ee instead of simply laudate.ee?

Both FQDN resolve to the same IP address because we configured a wildcard in the zone file.

Note that the same is true for mail.lino-framework.org, which is given as the Reverse DNS for the IP address of LF:

$ telnet mail.lino-framework.org 465

In the ovh console I noticed that the “Reverse DNS” for the LF server was set to “mail.lino-framework.org”. I changed this to “lino-framework.org” (without the “mail” subdomain). Afterwards everything (both sending and receiving a mail via luc@LF using TB) was still working.

It seems that the mail subdomain (or sometimes smtp or mx) is general practice. It makes sense to have your mail server on a different machine than your application. Already for security reasons. Also in order to be scalable.

https://serverfault.com/questions/711600/reverse-dns-is-not-a-valid-hostname-error-from-mxtoolbox

The free “MX plan” at OVH is limited to about 100 messages per hour. With our playing around today this limit has been reached and they blocked the account.

The mail log says:

to=<luc.saffre@gmx.net>, relay=ssl0.ovh.net[193.70.18.144]:587, status=deferred (SASL authentication failed; server ssl0.ovh.net[193.70.18.144] said: 535 5.7.1 Authentication failed)

It’s relatively easy to unblock it by simply changing the password. Though that means that we need to change the /etc/postfix/sasl_passwd file on every server that uses this account. That’s a disadvantage of using a relay host.

Rather than unlocking it I’d prefer to try once more the direct way of talking to the SMTP servers.

How to change your hostname:

$ sudo hostnamectl set-hostname laudate

But echo "body" | mail -s "some test" luc.saffre@gmx.net does not seem to understand the new hostname. It is still using the original hostname hetz1, which I specified at creation and which we changed to jane last week. Yes, I can say -a "From:me@my.com", but where does the mail command get the default value for the From: header field?

http://mailutils.org/manual/html_node/configuration.html

I can say:

$ mail --show-config-options | grep SYSCONFDIR
SYSCONFDIR=/etc       - System configuration directory

Which means that actually the config files are in /etc/main. And one of them, /etc/mail/local-host-names contains my default From header.

/etc/mail/local-host-names

$ mail --config-lint
mail: opening configuration file /etc/mailutils.conf
mail: configuration file /etc/mailutils.conf doesn't exist
mail: opening configuration file /home/luc/.mail
mail: configuration file /home/luc/.mail doesn't exist

Also I cannot find how to disable the relay host.

sudo postconf -d | grep relayhost

Jun 10 12:33:19 hetz1 postfix/smtp[16646]: A8CD040107: to=<tonis.piip@gmail.com>, relay=ssl0.ovh.net[193.70.18.144]:587, delay=1.8, delays=0.03/0.03/1.5/0.24, dsn=2.0.0, stat us=sent (250 2.0.0 Ok: queued as 404C2133B946F)

Jun 10 15:50:13 localhost postfix/pickup[2096]: 5DE2A7D07: uid=1001 from=<luc@localhost>
Jun 10 15:50:13 localhost postfix/cleanup[2124]: 5DE2A7D07: message-id=<20200610155013.5DE2A7D07@lino-framework.org>
Jun 10 15:50:13 localhost postfix/qmgr[17652]: 5DE2A7D07: from=<luc@localhost.org>, size=334, nrcpt=1 (queue active)
Jun 10 15:50:13 localhost postfix/smtp[2126]: 5DE2A7D07: to=<luc.saffre@gmx.net>, relay=mx00.emig.gmx.net[212.227.15.9]:25, delay=0.55, delays=0.05/0.05/0.26/0.19, dsn=5.0.0, status=bounced (host mx00.emig.gmx.net[212.227.15.9] said: 550-Requested action not taken: mailbox unavailable 550 Sender address has null MX (in reply to MAIL FROM command))
Jun 10 15:50:13 localhost postfix/cleanup[2124]: E4AAEA46B: message-id=<20200610155013.E4AAEA46B@lino-framework.org>
Jun 10 15:50:13 localhost postfix/qmgr[17652]: E4AAEA46B: from=<>, size=2350, nrcpt=1 (queue active)
Jun 10 15:50:13 localhost postfix/bounce[2127]: 5DE2A7D07: sender non-delivery notification: E4AAEA46B
Jun 10 15:50:13 localhost postfix/qmgr[17652]: 5DE2A7D07: removed
Jun 10 15:50:14 localhost postfix/local[2128]: E4AAEA46B: to=<luc@localhost.org>, relay=local, delay=0.08, delays=0/0.05/0/0.02, dsn=2.0.0, status=sent (delivered to command: procmail -a "$EXTENSION")
Jun 10 15:50:14 localhost postfix/qmgr[17652]: E4AAEA46B: removed

https://www.cyberciti.biz/faq/how-to-configure-postfix-relayhost-smarthost-to-send-email-using-an-external-smptd/

Cannot query “Eleonora …”: Must be “Client” instance

Tonis and I also had a look at #3683 (Cannot query “Eleonora …”: Must be “Client” instance.)

The problem is reproducible in a doctest snippet:

>>> import lino
>>> lino.startup('lino_book.projects.avanti1.settings.doctests')
>>> from lino.api.doctest import *
>>> u = rt.login('robin').get_user()
>>> qs = rt.models.uploads.Upload.objects.all()
>>> qs = qs.filter(project__coachings_by_client__user=u)
Traceback (most recent call last):
...
ValueError: Cannot query "Robin Rood": Must be "Client" instance.

Tonis observed that the problem “disappears” in the web client after the first request. That is, when you invoke the same menu command a second time (within a same runserver session), then the problem no longer occurs.

In our doctest the problem does not disappear after the first request:

>>> qs = rt.models.uploads.Upload.objects.all()
>>> qs = qs.filter(project__coachings_by_client__user=u)

Which means that something else makes it disappear… mysterious!