I've had enough of the out of control backup processes

[SacAutos]

I recently updated one of my servers from Plesk 11.0 to 11.5 . I was anxious to do so because I couldn't wait to start running monthly backups again. I had halted them because they ate up so much of the system's resources that apache was flapping madly, causing gateway timeouts throughout the backup period.

So this morning I went into the admin panel and reactivated the monthly backup with a scheduled start time of this Saturday at 9:30pm - a period of little usage. Yet, now I find that within a half an hour of enabling the monthly backup, one has just started. WTF! This is in the middle of the day - precisely NOT when I wanted this to run. Now I have to field complaints from customers as to why their websites are down with 504 Gateway Time-out messages.

Unfortunately, I've seen this behavior of the backups before on other versions of the panel on other servers, so I suppose I should have known better. BUT THIS IS CLEARLY BROKEN! Since nobody seems to have pointed this out - let me do so now. Backups are an essential part of running an IT operation. FIX THIS!!

 

[Linulex]

..... reactivated the monthly backup .... Backups are an essential part of running an IT operation. FIX THIS!!

I agree that backups are an essential part of any hosting company, but monthly backup ??? are you serious ????

We have never used the plesk backup and we use plesk sinds version 1.3.

We use rsnapshot. We retain 30 days worth of backup for every server and a backup takes from 10 to 30 minutes depending on how many domains there are on a server.

http://www.rsnapshot.org/

regards
Jan

 

[SacAutos]

I run backups on my websites daily, weekly, and now again, monthly. Not everything gets captured on the higher frequency ones - just the databases and places where files change a lot. This is the policy that we've established for this particular customer and circumstance. Your have your own criteria I'm sure.

My point is that Parallels appears to be unaware that their backup software is broken. It should work correctly as it is included with the software. Why should I have to pay for something else? If they can't get it to work correctly, they should pull it. Sorry about the harsh language but I'm feeling pretty P.O.'d that after all the feedback I've given Parallels in the past about their backup product nothing seems to have changed. Having pigz implemented has helped but it's still inadequate in the current state.

 

[AmeliaG]

My point is that Parallels appears to be unaware that their backup software is broken.

They are aware, there are many complaints about this bug and no solution yet!!

 

[Linulex]

Why should I have to pay for something else?

rsnapshot is open source. All it takes to implement is some of your time. In return you get:

- very fast backups
- more backups then plesk can provide

We use munin and mrtg on servers to check the status of the ventilators, the temperature of the cpu, i/o per disk, i-nodes used, etc... the usual stuff one should monitor. We also use installatron script installer because the built in as a laugh. rsnapshot you configure yourself, it backs up whatever you want, so also the munin config, the mrtg config, the installatron config, the mysql config, http config, and every other config you addapt yourself.

It should work correctly as it is included with the software.

I agree it should, but i also think it shouldnt be in there in the first place. plesk should be a hosting control panel and not becoming something that tries to take over the complete server.

regards
Jan

 

[IgorG]

FIX THIS!!

Fix what? Where is detailed report with logs, steps-to-reproduce, error messages and so on according to http://forum.parallels.com/showthread.php?t=106113 ?
How we can understand what exactly "DOES NOT WORK"?
Guys, complaints like "IT DOES NOT WORK! FIX IT!" is really not enough.
Thank you for understanding.

 

[AmeliaG]

Fix what? Where is detailed report with logs, steps-to-reproduce, error messages and so on according to http://forum.parallels.com/showthread.php?t=106113 ?
How we can understand what exactly "DOES NOT WORK"?
Guys, complaints like "IT DOES NOT WORK! FIX IT!" is really not enough.
Thank you for understanding.

Dear IgorG. I'm sure at this point Parallels know exactly what is going on with this backup bug! We understand but I really think it's time to work on this problem. So please ask support to read this forum sometimes.

Few day ago you asked me to do some tests with "curl FTP" , I did it and you never reply to me about those tests.

Thanks

 

[Juan ManuelT]

Fix what? Where is detailed report with logs, steps-to-reproduce, error messages and so on according to http://forum.parallels.com/showthread.php?t=106113 ?
How we can understand what exactly "DOES NOT WORK"?
Guys, complaints like "IT DOES NOT WORK! FIX IT!" is really not enough.
Thank you for understanding.

If you shedule a new backup process it starts few minutes later, without regard to the time you configure. Afterwards it starts working correctly. This has been happening in plesk 9, 10 and 11 and I dont think this is the first thread I see about it, I always though it was a known bug. Actually, I dont think I have ever been able to shedule a new backup without it happening. Ever. So it's easy to reproduce. Next time I change the schedule I'll save the logs and start a proper thread. This has happened to me in 2 separate instalations, the most recent I think it's even the ISO with the preinstalled plesk you could download at parallels of plesk 11 little after it was released, so I doubt it's a problem exclusive to my machines.

Both machines are installed in Ubuntu Server 10.04, as I said in versions 9 to 11 in several MUs. It's usually not noticeable if your server is not under load, it simply finishes the backup and then works correctly, but if your server is leaded it obviously slows everything down.

 

[SacAutos]

Juan has precisely described what has happened, and has been happening, in my environment as well. Schedule a backup to begin at some future date and watch it launch one almost immediately regardless of what time you've selected. We run CentOS for our servers (I have multiple) and have too witnessed this on versions 9, 10 and 11 of Plesk many times. Considering the lengthy Plesk 11.5 development cycle I'm shocked that this hasn't been addressed. I watched my server struggle for six hours yesterday through the heart of the business day as the unwanted backup process hammered it. Igor, I appreciate all that you try to do for the customers. However, this apparently an easy to reproduce issue and detailed logs aren't really necessary in this case.

 

[IgorG]

Guys, you should know that Scheduled backups on Plesk works as follows:

- schedule of running backupmng utility is defined in /etc/cron.d/plesk-backup-manager-task file (Once every 15 minutes. On each Plesk server specific moment chosen randomly)
- backupmng checks task records in psa.BackupsScheduled table and start them if they are exists.

As you can see delay may be not more than 15 min on Linux. On Windows it works at the exact time.

 

[Juan ManuelT]

Guys, you should know that Scheduled backups on Plesk works as follows:

- schedule of running backupmng utility is defined in /etc/cron.d/plesk-backup-manager-task file (Once every 15 minutes. On each Plesk server specific moment chosen randomly)
- backupmng checks task records in psa.BackupsScheduled table and start them if they are exists.

As you can see delay may be not more than 15 min on Linux. On Windows it works at the exact time.

I think you missunderstood, a 15 minutes delay is not the problem:

1) Assume it's, say, 1:49 PM

2)I login into the panel and configure the sheduled backups. I set it to daily and the time of execution to 7:49PM, so the backup should start in 6 hours

3) Backup starts a few minutes after instead, at 1:55PM or so (from what you've said, it starts the next time the backupmng utility starts via the cron)

4) Backup starts again at 7:49PM, and it works correclty from then on, starting at 7.49PM each day.

Unless I'm missunderstanding something here, step 3 shouldn't happen. Specially since it usually mean that backup starts during work hours, which usually are of heavier use. A warning from the interface would be cool too if this is intended.

 

[AmeliaG]

Hope we don't have to wait too long for Plesk #MU4

 

[SacAutos]

Deployed a new server on CentOS 6.4 with Plesk 11.5 . So far I'm not seeing the out-of-control issue occur any longer. Was there a fix or am I just kidding myself?

** update **

Perhaps I jumped the gun on this. Scheduled a bunch of backups to run on various evenings throughout the week. None have started so far. Will investigate further and report back later...

 

[SacAutos]

Now I see what's happening. I scheduled a bunch of backups to happen various days of the week in the evening of July 30th. (I did it in the evening in case the backups took off uncontrollably - at least they would have run at night.) However, none of the ran at all - even on the upcoming days that I scheduled them.

Instead, all the backups that I had scheduled for the various days of the week all ran in the wee hours of the morning of August 6th. That's exactly a week ago from the day I scheduled them.

Just to be clear, here's an example: I scheduled a backup for a customer account to run each week on a Thursday. The day I scheduled it was a Tuesday. The Thursday came and went without a backup running. Then Tuesday came along again and the backup ran.

As I said in the start of this thread, THIS IS CLEARLY BROKEN! Can somebody please fix this??