Resolved DNS Syncing from Plesk Obsidian to Digital Ocean corrupting today with domain name listed twice.

[famous]

Hello Plesk Gurus,

My first post here in Plesk Forums. Have patience.

I've used Plesk in a limited fashion over the years, but this week dove in big-time. I host dozens of websites for clients on digitalocean droplets running WordPress. We run one droplet for each client's website. Migrating some clients to shared environments using Plesk servers.

We're reasonably experienced with DNS and have been manually building DNS records for many years at Digital Ocean and elsewhere. This week as we've been migrating a few websites into Plesk and we've enjoyed having Plesk automatically update/build those DNS records with the Digital Ocean DNS Extension.

Everything was working perfectly with this automatic DNS record creation system. We migrated multiple sites from individual Ubuntu servers into this Plesk server. Loving Plesk... Until this morning.

We started another migration and noticed the DNS for the migrated site was not propagating as the prior 6 or 7 sites did nearly instantly after Plesk updated Digital Ocean DNS. A closer examination of that domain's DNS at Digital Ocean showed corruption.

PROBLEM: When our Plesk server syncs DNS records to Digital Ocean, our individual record entries are corrupted with the domain name repeated. It is doing this on every DNS record including A, CNAME, MX, TXT records.

EXAMPLE: For a given A record that should simply show example.com pointing to the machine IP
How it should be: somedomain.com.
How it is with this bug: somedomain.com.somedomain.com.

So basically on every DNS entry where it should be the domain name listed once... it is listing it twice.

DETAILS:

• Plesk Obsidian Version 18.0.42 Update #1
• Ubuntu 20.04.4 LTS
• Digital Ocean dedicated CPU droplet
• Convinced that things were going well, we upgraded from a free Plesk license included with Digital Ocean's Plesk marketplace box to Plesk Web Pro Edition yesterday so we could import more than 3 websites.

THINGS WE'VE TRIED:

1. Changed the domain name inside of plesk from example.com to example.net to let it sync to DigitalOcean. The sync does happen but the domain is still listed twice: example.net.example.net.
2. Risked corrupting another domain for a rarely used website. (This particular domain was migrated into Plesk yesterday and Plesk CORRECTLY built the DNS records yesterday.) First double-checked that the DNS records over at Digital Ocean look good. Then adjusted the domain name in Plesk and watched how the sync looked... yup. Corrupted at Digital Ocean. So the problem doesn't seem to be with the first domain's Plesk records.
3. Reboot Plesk server. More testing. Corruption still there.
4. Opened ticket much earlier today with Plesk Support. They've not gotten to it yet...

Anyone have experience with this issue?

Thank you!
Clark

 

[motom001]

Hello,

I can confirm this bug since yesterday evening while refresh the Let's Encrypt certificates.

Details:
CentOS Linux 7.9.2009 (Core)
Plesk Obsidian Version 18.0.42 Update #1
DigitalOcean DNS in version 1.3.2-119

First time bug seen: Wed, 06 Apr 2022 19:56:11 +0200

This bug is high critical and I have no workaround at the moment. When refreshing an domain dns record (e.g. for Let's Encrypt) will break the dns config. So each domain will break up in the next 90 days.

 

[Kaspar]

Sorry to hear the DNS synchronization with DO got broken somehow. I just spun up a Droplet with Plesk and installed the DO DNS extension to test the issue you're facing. I can confirm that for every domain I add in Plesk the DNS records get screwup, listing the domain twice in the host of all records. It's strange that this started yesterday as the changelog doesn't list any updates for Plesk or the DO DNS extension.

Good thing you've opent a ticket with Plesk support. Hopefully they can get this issue sorted. Let us know if they've been able to help you out.

You might be best off disabling the DO DNS extension and managing the DNS manually for now.

 

[TomBoB]

Yep, can also confirm the same happening on all our servers.

It appears someone forgot to put the famous . at the end of the records. All records done have just the proper record as it should be, but its hostname another time added to it. Happens when the final . isn't behind the hostname.

 

[TomBoB]

from Plesks facebook group just now:

„The behavior in question was registered as a Plesk bug: #EXTPLESK-3432“
„The bug fix is expected in the next extension's update: DigitalOcean DNS 1.3.3“

 

[Tim den Dikken]

from Plesks facebook group just now:

„The behavior in question was registered as a Plesk bug: #EXTPLESK-3432“
„The bug fix is expected in the next extension's update: DigitalOcean DNS 1.3.3“

Yes, it's definitely a bug. I just discovered it after adding a subdomain to my server. When you add a record it does an automatic re-sync. I worked around the bug exporting the DNS zone from Plesk and disabling the DNS sync for now. I manually added the records that were in the Plesk zone to the DigitalOcean DNS in the customer backend.

 

[famous]

Sorry to hear the DNS synchronization with DO got broken somehow. I just spun up a Droplet with Plesk and installed the DO DNS extension to test the issue you're facing. I can confirm that for every domain I add in Plesk the DNS records get screwup, listing the domain twice in the host of all records. It's strange that this started yesterday as the changelog doesn't list any updates for Plesk or the DO DNS extension.

Good thing you've opent a ticket with Plesk support. Hopefully they can get this issue sorted. Let us know if they've been able to help you out.

You might be best off disabling the DO DNS extension and managing the DNS manually for now.

Thank you. Yes, that is my present work-around. We have manually fixed the corrupted DNS records over at DO.

 

[famous]

from Plesks facebook group just now:

„The behavior in question was registered as a Plesk bug: #EXTPLESK-3432“
„The bug fix is expected in the next extension's update: DigitalOcean DNS 1.3.3“

Thank you for letting us know.

How long does it typically take Plesk Dev to release bug fixes?

 

[famous]

A recommended way to disable Digital Ocean DNS Extension within Plesk: Revoke API or OAuth2 authorization.

Do NOT use the "Deactive" button within the extension. Here's why:

As we were testing the Digital Ocean Extension on some test sites early this week in anticipation of bringing client websites to Plesk, we realized that seemingly normal behavior is that if you deactivate the DO extension, it sometimes goes in and deletes ALL DNS entries at Digital Ocean. We saw this happen more than once.

I mention this because if you decide to deactivate the Digital Ocean Extension (boy I would if I were you), you might want to just go into the extension and revoke your API or OAuth2 authorization... which will cut off the extension's access to your Digital Ocean account without giving it a chance to delete good records. If you host a lot of sites, this can represent a MASSIVE amount of work to recreate dozens and dozens of DNS records.

Once the bug is fixed, it will be simple to re-authorize the DO extension.

 

[famous]

In the ticket we opened with Plesk Support on this issue yesterday we were told:

We have already escalated this issue to our development team. They are doing their best in order to sort it out. Once we receive any news from them, I will let you know immediately.

If we get any updates, we'll post them here.

 

[TomBoB]

A recommended way to disable Digital Ocean DNS Extension within Plesk: Revoke API or OAuth2 authorization.

Do NOT use the "Deactive" button within the extension. Here's why:

As we were testing the Digital Ocean Extension on some test sites early this week in anticipation of bringing client websites to Plesk, we realized that seemingly normal behavior is that if you deactivate the DO extension, it sometimes goes in and deletes ALL DNS entries at Digital Ocean. We saw this happen more than once.

I mention this because if you decide to deactivate the Digital Ocean Extension (boy I would if I were you), you might want to just go into the extension and revoke your API or OAuth2 authorization... which will cut off the extension's access to your Digital Ocean account without giving it a chance to delete good records. If you host a lot of sites, this can represent a MASSIVE amount of work to recreate dozens and dozens of DNS records.

Once the bug is fixed, it will be simple to re-authorize the DO extension.

Interesting. Have just double-checked that. Obviously want to avoid that scenario. But in our case your scenario hasn't happened to a single domain. May be it was an additional glitch your side. Deactivating worked flawlessly for us.

 

[famous]

Good news.... a fix is in the works and expected tomorrow.

From Plesk Support on the ticket we have open for this bug:

The developers are doing their best to release a fix to this bug. It was caused by the unexpected changes in the DigitalOcean API. They are planning to deliver the fix tomorrow in version 1.3.3 of the DigitalOcean DNS extension.

 

[famous]

Interesting. Have just double-checked that. Obviously want to avoid that scenario. But in our case your scenario hasn't happened to a single domain. May be it was an additional glitch your side. Deactivating worked flawlessly for us.

I'm glad it did not bite you! We definitely saw that behavior. It did not happen every time but it did happen. We tested the extension thoroughly before migrating live customers over to Plesk. Initially, we didn't think we would like it and intended to continue manually DNS record entries, but it won us over.

The extension is an awesome time-saver. It creates a few records differently than how we were doing it manually, but we figure it probably is doing a better job with better "best practices" than what we used. All in all... we love it.

 

[Manuel Sosi]

Goodmorning everyone, do you have news about this version 1.3.3? On my plesk server it still shows 1.3.2 available ... and of course I have to manage all domains manually.

Oh and btw thank you famous for reporting the issue.

 

[motom001]

And now, ladies and gentlemans, nothing changed and everything works again...
I have still the DO version 1.3.2-119 but the entries on DO side are all correct.
Maybe has DO rollback an braking API-update?
Could somebody please inform the plesk developers about the chaning frame conditions?

 

[Manuel Sosi]

I confirm that!

Thank you for the notice, I'm reactivating all the DO DNS extensions now

 

[famous]

The ticket I had open for this with Plesk Support changed to RESOLVED an hour ago. Perhaps this was resolved on the Digital Ocean side as the devs from Plesk worked with them on it?

At any rate, I'm going to turn my DO DNS Extension back on right now based on motom001 and Manuel Sosi's experiences.

 

[famous]

RESOLVED.

I just put my API Key back into the Digital Ocean DNS Extension and let Plesk update a domain I had been testing with when I discovered this corruption.

Indeed it appears this bug has been fixed... apparently on the Digital Ocean side. The record corruption is gone.

 

[famous]

This fix has just been confirmed in my ticket with Plesk Support:

I could confirm that the issue was fixed from Digital Ocean side.
Our developers will continue working in collaboration with Digital Ocean developers in order to avoid such issues in future.

 

[motom001]

Great work and thx to the plesk team, especially the developers!