Server struggling over last hour

[whole diary]

[grahame]

All go (or all stop) at the moment.   What looks like a denial of service attack just blocked - server has been severely loaded for the last hour or so due to what looks like an out or control spider.

Not totally sure all is fixed - this is something of an experimental post

[ray951]

I am not sure that this is related but there seems to be an issue with the 'Most liked recent subjects' column which is about 1 inch wide on my PC (Firefox) and therefore the 'Hello' section is taking up the whole screen.

[grahame]

I am not sure that this is related but there seems to be an issue with the 'Most liked recent subjects' column which is about 1 inch wide on my PC (Firefox) and therefore the 'Hello' section is taking up the whole screen.

Not the same issue .... thanks for noting the Direfox issue ... I think I have that fixed now, hopefully not breaking anything else / anyone else's browsers.

[ray951]

Thank, now works fine for me.

[Bmblbzzz]

There seemed to be a problem with the site all this afternoon (2nd May). Couldn't reach it at all; other sites working fine.

[grahame]

There seemed to be a problem with the site all this afternoon (2nd May). Couldn't reach it at all; other sites working fine.

Odd.  No other reports, and server loading graphs don't show any deviation from normal levels. Potentially a blip somewhere at a DNS^▸ (Domain name service) server or routing table somewhere??

[grahame]

All go (or all stop) at the moment.   What looks like a denial of service attack just blocked - server has been severely loaded for the last hour or so due to what looks like an out or control spider.

Not totally sure all is fixed - this is something of an experimental post

And again ... 6,000 requests in around 8 minutes (14:21 to 14:29) from an IP address that's pointed to New Jersey. Server load normal "1" or "2" (I can explain units at some time) shot up to 84.46 and the server emailed me in panic.  Back down in normal range now.

Sorry if anyone had any problems ... this is a test post.

[TonyK]

And again ... 6,000 requests in around 8 minutes (14:21 to 14:29) from an IP address that's pointed to New Jersey. Server load normal "1" or "2" (I can explain units at some time) shot up to 84.46 and the server emailed me in panic.  Back down in normal range now.

Sorry if anyone had any problems ... this is a test post.

Hadn't noticed anything. It seems you are a lot better prepared than the NHS or Travelwest, grahame!

[grahame]

And again ... 6,000 requests in around 8 minutes (14:21 to 14:29) from an IP address that's pointed to New Jersey. Server load normal "1" or "2" (I can explain units at some time) shot up to 84.46 and the server emailed me in panic.  Back down in normal range now.

Sorry if anyone had any problems ... this is a test post.

Hadn't noticed anything. It seems you are a lot better prepared than the NHS or Travelwest, grahame!

Should be ... I (used to) teach server setup and admin.  Mind you, the server software is designed not to actually fall over if it gets hit by what is/was essentially a denial of service attack, and that helps avoid a total failure. We have enough memory /swap space to avoid overflowing too.  I suspect that the "attack" wasn't malicious but someone's test robot /spider which was incorrectly tuned.

Attachment shows the graph of server load over recent days and today's spike is obvious.   The figure is a weighed average for the last 15 minutes - thus the apparent slow recovery.  You can spot major backup cycles in there too.