363 reputation
15
bio website
location
age
visits member for 1 year, 10 months
seen Jul 17 at 3:44

Jul
17
awarded  Excavator
Jul
17
awarded  Editor
Jul
17
revised Copy data through SSH tunnel over multiple hops
tip to silence the annoying errors
Jul
17
suggested approved edit on Copy data through SSH tunnel over multiple hops
Feb
14
awarded  Yearling
Sep
25
comment Forensic Analysis of the OOM-Killer
The OOM Killer log isn't really that helpful. The problem is that Linux over-commits memory. This means that a random process will get an error when it tries to use previously allocated memory. If you turn off overcommit, then you will get an error right when an application requests too much memory instead of much later and in a different process.
Sep
24
comment Forensic Analysis of the OOM-Killer
Install atop. It will write an overview of system activity every 10 minutes. From that, you can easily detect slowly growing processes, etc. It won't always catch a fast culprit, but it has helped me several times.
Sep
24
answered HAProxy not balancing server load fairly
Jul
19
comment Is My Website Under DDOS Attack?
Sorry, I did assume you were using PHP in Apache. If you're using Nginx, you're fine. You need to use "limit_req_zone" in nginx instead of mod_evasive on Apache.
Jul
6
answered Is My Website Under DDOS Attack?
Jun
15
comment Simple upstart script for nodejs + forever on Ubuntu
My hack is for logging to arbitrary files. The console setting is for logging to the Linux console. This might be useful if your console goes to syslog and that's really where you want your logs.
May
29
comment Simple upstart script for nodejs + forever on Ubuntu
Please don't move the goalposts after I answered your original question.
May
18
answered Simple upstart script for nodejs + forever on Ubuntu
May
18
comment Nginx file redirect
location ~ ^/images { proxy_cache_key $scheme$proxy_host$uri; } # don't use $args when checking cache.
May
11
comment haproxy and keepalived on Amazon EC2
+1. EIP transfers take 10-30 seconds and instantly notify all current clients (via TCP reset). ELB changes take minutes (or longer because not everybody respects DNS), and clients must timeout before noticing that the load balancer has died.
May
11
comment Nginx file redirect
If it's about caching, you could have /images use a different cache-key (without the query args).
May
11
answered TCP vs UDP Error-Checking
May
11
awarded  Critic
May
11
comment Amazon EC2 Instances
If the original sever won't boot, cloning may not help. It may be best to launch a new server from the base OS AMI, then attach the old drive and copy the data off.
Feb
19
awarded  Supporter