/save-off (Big issue)

[mrchasez]

This is a copy and paste from a ticket, which was requested to paste here.
For a bit now, i have been trying to find the issue with our chunks. It seemed as though after being up for a bit, chunks just wouldn't unload and i couldn't find out why.

Well... i found out why...
At some time, or a few. The server runs /save-off (and not /save-on)
Which stops all chunks from unloading. I even posted on the spigot forums, which is where i found out about this.

So i went ingame and our TPS was horrible, because we had 55,000 chunks loaded. So i typed /save-on and within thirty seconds it went down to 5000 normal chunks. So there must be some script turning /save-off on for some reason, how do i stop it from ever running! (Not /save-all)

[.=QUACK=.Major.Pain]

Have you looked in your server.properties?

[Edge100x]

On our end, we run save-off, then save-all, then save-on every hour when we make snapshots. save-on is definitely part of this process, but the system won't proceed with save-all or save-on unless save-off returns a valid response -- these three are accepted:

Disabling level saving
Turned off world auto-saving
level saving

What does your server return when you run "save-off" by hand? It is possible that one of your plugins is changing its behavior.

[mrchasez]

On our end, we run save-off, then save-all, then save-on every hour when we make snapshots. save-on is definitely part of this process, but the system won't proceed with save-all or save-on unless save-off returns a valid response -- these three are accepted:

Disabling level saving
Turned off world auto-saving
level saving

What does your server return when you run "save-off" by hand? It is possible that one of your plugins is changing its behavior.

It seemingly works fine. However i just logged in and we had 20,000 chunks loaded.
So i ran /save-on and it went down to 6,000 almost instantly.

How can i disable part of the script. I don't know why it does more then /save-all anyway.
However the /save-off is ruining gameplay. Also, why can't the script save one minute before the hour or something. Because it saves minutes before the hour, and we restart ON the hour (since its the only time you can) rollbacks can occur. However fixing the save-off is something we need done priority.

[Edge100x]

It seemingly works fine. However i just logged in and we had 20,000 chunks loaded.
So i ran /save-on and it went down to 6,000 almost instantly.

You actually seem to be saying that it is not working. In that case, I need to know how your server is responding to "save-off". Many plugins modify standard game responses and that's likely just what's happening to you.

Please let me know what your server is saying in response to "save-off". Alternately, open a support request on this so that I can run "save-off" for you and obtain the response directly.

How can i disable part of the script. I don't know why it does more then /save-all anyway.

It is necessary to turn off saves so that the server does not write to disk while the snapshot is being made (this is why the command exists). When it writes to disk while the snapshot is in progress, the snapshot becomes corrupted.

Also, why can't the script save one minute before the hour or something. Because it saves minutes before the hour, and we restart ON the hour

It does not save exactly on the hour. Each server is done serially a bit a little after the top of the hour.

[mrchasez]

-

I have a ticket with more info, i asked if it could be forwarded to you.

[Edge100x]

I have a ticket with more info, i asked if it could be forwarded to you.

I thoroughly responded to your support request a few hours before this post from you here.

I found that your server was responding to "save-off" correctly, and that our system was correctly performing "save-all" and then "save-on". Each of your servers' logfiles showed that everything had been working correctly every hour since the last time you wiped the log.

This suggests that you may have had a plugin problem that has since been resolved, or that something else is going on, independent of our system's behavior.

Also keep in mind that Minecraft doesn't save continuously any more. It saves something like every 15 minutes. It is possible that you just triggered an extra save operation, and in that case, if you saw an improvement from the extra save, you might consider running a plugin that causes saves to be performed more frequently.

[mrchasez]

I have a ticket with more info, i asked if it could be forwarded to you.

I thoroughly responded to your support request a few hours before this post from you here.

I found that your server was responding to "save-off" correctly, and that our system was correctly performing "save-all" and then "save-on". Each of your servers' logfiles showed that everything had been working correctly every hour since the last time you wiped the log.

This suggests that you may have had a plugin problem that has since been resolved, or that something else is going on, independent of our system's behavior.

Also keep in mind that Minecraft doesn't save continuously any more. It saves something like every 15 minutes. It is possible that you just triggered an extra save operation, and in that case, if you saw an improvement from the extra save, you might consider running a plugin that causes saves to be performed more frequently.

No plugins have changed since this update. However, a version of spigot did change (a few).
However, spigot don't run save off. Only your system does. On all the hosts i have been with none have ever ran /save-off before. You say it has something to do with backups? Sounds like that is the issue at hand. There isn't anything we can do as nothing saves but your script.

Ill try to see if we can keep going, but whenever this issue strikes, it totally kills performance and
gameplay because of those chunks.

[Edge100x]

Yes, as I mentioned, save-off is necessary to make external backups. Any host that makes external backups that is not using save-off/save-all/save-on would be doing you a disservice, as the backups could be corrupted. Again, that is why these commands exist.

Not a lot of hosts even make snapshots or backups. We do, and we do it right.

As I mentioned, the system has behaving correctly as far back as I could look in your logs. This means that you are not seeing a problem right now, unless you have a plugin that is keeping them turned off when we tell the server to turn them back on.

If you truly don't believe me and want to just shut everything off, you can also do that. Turn off snapshots entirely, and the system won't need to follow the procedure every hour.