[Solved] Very busy today?

[Andyroo]

I’m getting the ‘Sorry we are busy’ page nearly every alternate time I try to read or post but:

1) There are no cloudflare errors reported
2) There does not seem to be that many posts being made (using unread menu option)
3) Response for the article is fine either it’s a100% fail or fast delivery.

[rpdom]

Yes, I've been getting a few of those too for the last couple of days.

[pcmanbob]

First started getting the busy message Friday afternoon , Saturday morning was not to bad but Saturday afternoon it was happening so often I logged off .
Today (Sunday ) has not been so bad , with little or no busy messages this evening.

May just a lot of visitors to the forum but not many posting.

[Andyroo]

Seems to be getting worse again

[jamesh]

In an peak of irony seldom reached, as I clicked on this thread title to check it out, I got the Busy Page...

Anyway, it's been reported upstream.

[Andyroo]

Ta muchly

[hippy]

Just to note that things seemed to have settled down from a week ago but I have been getting a lot of 'high traffic' pages throughout today.

[jamesh]

Indeed, looking slow again. Probably the number of bannings that have had to be done today (lots of spam). If a ban hasn't been done for a while, database caches need to be refreshed, and that takes ages, meanwhile the rest of the forum stops dead while that is done.

It's a bit pants, but it's the way PHP is architected.

[mahjongg]

I'm trying to avoid unnecessary banning.
banning each and every spammer is counter productive, only if he/she is still busy with a stream of spams I try to stop him with a ban.
Otherwise banning spammers is just counter productive, they just invent a new user name with a push of a button, (or just automatically in case of spam bots).

[LTolledo]

...am getting more "hits" today than before....

well if its a necessary "housekeeping" routine... then me just endure it ....

as the site is accessed from everywhere in the world there is literally no "night shift cleaning window" so somewhere, sometime, someone gets hit by the "busy banner"....

[pcmanbob]

Seems particularly bad this evening. Often getting the busy page and a general slow loading of treads, with the waiting for raspberrypi.org message in Firefox.

[Douglas6]

Seems particularly bad this evening.

Ditto.

[Andyroo]

Gave up just after midday UK time - checked a couple of more times still 75% busy messages.

Even now just after midnight UK time as bad.

Edit:
6 try to post the original part of this message

[pcmanbob]

One thing I did notice yesterday, if I was logged in I was getting the busy page , but if I logged out , I did not and the forum seemed faster in loading pages.

Next time this starts I an going to test it using 2 devices one logged in and one not , just to make sure I was not imagining it.

[Andyroo]

That does not surprise me as the board keeps track of the messages you have read vs new postings so the database access is more for logged in accounts than general browsing.

Obviously any replies etc you make add to the index but on average I think many more database updates occur for logged in accounts compared to responses

[jamesh]

That does not surprise me as the board keeps track of the messages you have read vs new postings so the database access is more for logged in accounts than general browsing.

Obviously any replies etc you make add to the index but on average I think many more database updates occur for logged in accounts compared to responses

Exactly. I've reported the problem upstream, so hopefully Mythic will be able to take a look to see if there is an issue somewhere.

[Andyroo]

Seen the post about patches / fixes / mods going on and not seen any errors in the last few times I visited here.

Not on it that heavy (currently on the mobile in the library) but first impressions are that it’s lots better

[jamesh]

The patches were actually for another server (Blog wordpress server) , the problem with the forum was a bad network card apparently. All sorted now, please post here if problems still turn up.

[DougieLawson]

The patches were actually for another server (Blog wordpress server) , the problem with the forum was a bad network card apparently. All sorted now, please post here if problems still turn up.

Haven't seen a "we're busy" at all today. So looks like it's fixed from my point of view.

[Andyroo]

Been great now I’ve had a good ‘play’ around.

Thanks for pushing it up the line for us.

[Pete_Stevens]

Just a bit more detail. We had a bug in the module in the intel network driver (Intel 710) that comes with Debian Stretch. We built a replacement using the newer bugfixed sources direct from intel which we've seen from other machines fixes the issue.

The challenge is installing it, you have to drop the network on the server, remove the old module, install the new module and bring the network back up and we're remote from the server. If it goes wrong half way through the box id offline and you can't log in and fix it. Fortunately in Mythic we have serial as well as network on all our servers. We logged in over serial and then ran a script that completed all of these steps as quick as possible. As a result we interrupted network for around a second while the fix went in. Since then we've seen the load graph return to normal and the alerts have stopped.

[jamesh]

Just a bit more detail. We had a bug in the module in the intel network driver (Intel 710) that comes with Debian Stretch. We built a replacement using the newer bugfixed sources direct from intel which we've seen from other machines fixes the issue.

The challenge is installing it, you have to drop the network on the server, remove the old module, install the new module and bring the network back up and we're remote from the server. If it goes wrong half way through the box id offline and you can't log in and fix it. Fortunately in Mythic we have serial as well as network on all our servers. We logged in over serial and then ran a script that completed all of these steps as quick as possible. As a result we interrupted network for around a second while the fix went in. Since then we've seen the load graph return to normal and the alerts have stopped.

Thanks Pete, if it's OK with you'll i'll get back to writing the Pi Linux drivers, seems simpler!

[Heater]

Pete_Stevens,

Sounds like a neat bit of sleuthing and fixing. Great stuff.

I love little stories of tech battles like that. Thanks.

[Andyroo]

Pete_Stevens,

Sounds like a neat bit of sleuthing and fixing. Great stuff.

I love little stories of tech battles like that. Thanks.

+1 to that. Lost count the number of times I’ve had to go to site or get local support to push a button or complete a simple job...

Thanks to all for fixing it.