Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I have an instance running at Amazon EC2. I just checked the monitoring, and I saw that the server was down for a while.

TO be more precise, there is completely no line between 15:16:00 and 15:46:00 in the monitoring graphs of the EC2 console. I also have confirmation from Uptimerobot that my servers where down.

Apparently my server was down for exactly 30 minutes. I have gone trough nginx logs, and the system log, but I could not find anything out of the ordinary. Everything works just fine now.

Can I find out what happened somehow, it is really strange.

This is what happened to php-fpm.

[29-Dec-2011 23:27:34] NOTICE: fpm is running, pid 1131
[29-Dec-2011 23:27:34] NOTICE: ready to handle connections
[04-Jan-2012 15:48:07] NOTICE: fpm is running, pid 1169
[04-Jan-2012 15:48:07] NOTICE: ready to handle connections
[04-Jan-2012 15:51:22] NOTICE: fpm is running, pid 1167
[04-Jan-2012 15:51:22] NOTICE: ready to handle connections

Nginx log. There was no real activity during that period. The server is only used for a small website for now.

220.181.108.175 - - [04/Jan/2012:14:30:50 +0000] "GET / HTTP/1.1" 404 22 "-" "Mozilla/5.0 (compatible; Baiduspider/2.0; +http://www.baidu.com/search/spider.html)"
123.125.71.105 - - [04/Jan/2012:14:32:14 +0000] "GET / HTTP/1.1" 404 22 "-" "Mozilla/5.0 (compatible; Baiduspider/2.0; +http://www.baidu.com/search/spider.html)"
74.86.158.106 - - [04/Jan/2012:15:48:41 +0000] "GET / HTTP/1.1" 200 9208 "-" "Mozilla/5.0+(compatible; UptimeRobot/1.0; http://www.uptimerobot.com/)"
124.115.0.157 - - [04/Jan/2012:15:56:15 +0000] "GET / HTTP/1.1" 301 178 "-" "Sosospider+(+http://help.soso.com/webspider.htm)"
74.86.158.106 - - [04/Jan/2012:15:56:43 +0000] "GET / HTTP/1.1" 200 9208 "-" "Mozilla/5.0+(compatible; UptimeRobot/1.0; http://www.uptimerobot.com/)"
74.86.158.106 - - [04/Jan/2012:15:57:50 +0000] "GET / HTTP/1.1" 200 3836 "-" "Mozilla/5.0+(compatible; UptimeRobot/1.0; http://www.uptimerobot.com/)"
77.21.146.23 - - [04/Jan/2012:16:06:52 +0000] "GET /robots.txt HTTP/1.1" 301 178 "-" "findlinks/2.0.2 (+http://wortschatz.uni-leipzig.de/findlinks/)"

Further, the only ports that are open on the system are 80, 443, 12345 (ssh). I do not know where to find the actual ssh log, but I did a logwatch dump, and SSH showed nothing.

These are the monitoring graphs

enter image description here


@James Little

I have checked /var/log/btmp, the file has been last changed 1-1-2012 and is 0 bytes.

ifconfig show me everything 0, I assume no errors and everything is ok. I don't really have the knowledge to work with ifconfig and ethtool as you suggested. I tried some google searches but failed to find some solid methods that would give me some information.

I think I will send an email to Amazon now, maybe they have some answers.

share|improve this question
    
You say the nginx log looks normal; do you have access logging enabled? i.e. you can see accesses over the internet during this time? Also, did you notice what the load average was when you first re-connected to the instance? –  James Little Jan 4 '12 at 18:06
    
I have access logging enabled, and there was nothing during that time period. The time period just before downtime there was just normal behavior, no attacks or anything. When the server started again at 15:46:00 CPU usage was extremely high for 6 minutes, the highest it's ever been. Also Network utilization during these 6 minutes were just above avarge. But in that 6 minutes time perion, there is absolutely no activity in the nginx access logs. Ill edit in php-fpm's log, it went down, up and rebooted I guess. –  Saif Bechan Jan 4 '12 at 18:15
    
Have you checked your bad login attempt log (usually /var/log/btmp)? I think you will have to ask Amazon, maybe they detected an attack (e.g. syn flood) and took action higher up the chain to cut off the traffic. Also use ifconfig and ethtool to check for interface errors. –  James Little Jan 4 '12 at 18:37
    
Also, how long has the nginx process been running? You can check the start-time using ps aux or similar. –  James Little Jan 4 '12 at 18:43
    
I will check out the log, and try to check interface errors. I am new to all this, that is why I only run a small website on the server right now, I knew stuff like this would happen. Hope I can find some answers in some logs, I will also send an email to amazon. –  Saif Bechan Jan 4 '12 at 18:43
show 1 more comment

2 Answers 2

up vote 0 down vote accepted

You don't specify if it actually rebooted. In case you did not check - use uptime to see when it last rebooted, or go thorough syslog or dmesg (from your php-fpm log I guess it did reboot). Since it was unavailable for some 30 minutes, it doesn't looks like some planned upgrade (unless they decided to "update" all the datacenter instances at once ;) .

If it was reboot, it's either some failure inside your instance or failure at amazon - again, look at syslog/dmesg.

If it wasn't reboot it could be also some issue that affected just the monitoring.

Amazon have status page of their datacenter issues, with history (somewhere on your EC2 dashboard). For planned reboots, in EC2 you have history too (under EC2 it's just above "instances", if I remember well).

Single instance unavailability is a normal (I did not say common) issue though. It's not feasible to totally prevent it.

share|improve this answer
add comment

Was this downtime a couple of weeks ago? Amazon did a rolling upgrade-and-reboot of all EC2 instances in mid to late December. They emailed everyone about it a week or two before they did it.

share|improve this answer
    
No it was today actually, just a few hours ago. I live in the Netherlands, the times that I gave were from today here in the Netherlands, about 5 hours ago. I have also checked if there was any downtime, but I could not find any information. My instance is located in the region west-europe, ireland. –  Saif Bechan Jan 4 '12 at 17:51
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.