Reputation
468
Top tag
Next privilege 500 Rep.
Access review queues
Badges
2 10 29
Impact
~88k people reached

Nov
24
comment Where to start with diagnosis of dropped network connectivity on Debian?
Thanks. There are definitely no problematic gateways involved, as I have multiple servers on the same subnet (1 hop away from each other), and they cannot communicate with the problematic server when this problem occurs. So from your list I assume we can rule out gateway and network issues. That leaves VM->Host issues. Regarding DHCP, from our VM's point of view we are not using DHCP, but rather have a static IP. How these IPs are allocated by the host though I don't know - I suppose something similar to a DHCP lease could be used?
Nov
24
comment Where to start with diagnosis of dropped network connectivity on Debian?
@PaulHaldane - the only entries in messages around that time are from the reboot I did to fix it - nothing from the problem itself.
Nov
23
comment Where to start with diagnosis of dropped network connectivity on Debian?
These are virtual servers servers are hosted with a VPS provider, so the physical layer is not accessible to me. Also since this has happened on multiple servers, it's unlikely to be an NIC fault or cable etc.
Nov
5
comment Ulimit file descriptor limits not being applied for particular process
Thanks. It uses --chuid redis:redis on start-stop-daemon. I've added the ulimit to the startup script now.
Nov
5
comment Ulimit file descriptor limits not being applied for particular process
Great summary Kannan! I've now gone ahead and added the ulimit to the startup script. It's a pity there is no way of just setting the limit for the user which will work with deamonised processes (as we have multiple startup scripts), but this works. Thank you!
Nov
4
comment How to check current TCP Congestion Window on Linux?
Ah, indeed it was 10 :)
Nov
4
comment How to check current TCP Congestion Window on Linux?
Ok thanks - I'll try to get the source, although this is a custom build for a Xen VPS. A pity it's not exposed to the users. Thanks for providing a definitive answer though.
Nov
4
comment Ulimit file descriptor limits not being applied for particular process
Thanks. I suspected that wasn't quite right. The process is being started this line in an init.d script : if start-stop-daemon --start --quiet --umask 007 --pidfile $PIDFILE --chuid redis:redis --exec $DAEMON -- $DAEMON_ARGS. Which pam would be appropriate in this case?
Nov
4
comment Ulimit file descriptor limits not being applied for particular process
Thanks - I'll consider raising the fs.file-max value then. Regarding pam - I believe we are using this (I've added some extra config to the question). I'm not really sure I understand PAM though, as our settings seem to be relate to SSH, which is not used by this user. Is there any other file I need to configure? I could always put a ulimit setting in the init scripts for redis, but would rather not have to do this.
Nov
4
comment How to check current TCP Congestion Window on Linux?
Thanks. Still would like to be able to verify though.
Nov
4
comment How to check current TCP Congestion Window on Linux?
Yes, that only shows the value if you have changed it yourself unfortunately. I'm looking for a way of finding the current value regardless of whether or not it was changed.
Nov
4
comment linux initcwnd and initrwnd via /etc/sysctl.conf
What if its a different interface that you want to change the setting of, other than the default one?
Nov
3
comment Recommended TCP keepalive settings for a busy server
Hmm, not sure - I'll look into that, thanks!
Nov
3
comment Recommended TCP keepalive settings for a busy server
Redis holds all of its data in memory (we don't even have disk snapshotting enabled). Average response times from both servers are in the milliseconds. Yeah, there must be something odd going on, but not sure what yet.
Nov
3
comment Recommended TCP keepalive settings for a busy server
Specifically I'm seeing this from an app server communicating with a redis server. Neither server is stressed or responding slowly though.
Nov
3
comment Recommended TCP keepalive settings for a busy server
At the application level these errors are 'read ETIMEDOUT', but I'm not sure how I'd give you any lower-level info than that.
Nov
3
comment Recommended TCP keepalive settings for a busy server
Only timeout I can see is /proc/sys/net/ipv4/tcp_fin_timeout which is 60. Is that what you meant? (Just realised you might be talking about the timeout errors we have been getting - will try to get these for you)
Sep
30
comment How to reliable re-install nf_conntrack module after removal?
Hmm, lsmod returns nothing (and I haven't removed anything yet). Perhaps I'm misunderstanding something?
Sep
29
comment How to reliable re-install nf_conntrack module after removal?
@HBruijn - I realise that it's necessary for certain types of rules in iptables (like those using the state keyword), but as I understand it, if these rules are not being used, then it is unnecessary. Is there something non-iptables that uses it in a standard config?
Sep
29
comment HAProxy high CPU usage - configuration issue?
Splitting into two processes is an interesting idea that I had not considered, thank you. Static traffic is very low, as we are behind a CDN, however spitting http app calls and the websocket traffic might be worth taking a look at. I suppose it would complicate our DNS records a little bit, but not a big deal. I'd like to see if I'm doing anything terribly wrong with the config first, but will certainly consider this.