Linode.com slowness

@ryantate:

I am seeing some pretty severe slowness this morning, not just on my own Linode but even accessing www.linode.com. It has improved slightly but is still pretty slow.

I'm seeing the same problem, for both of my Linodes (which are in Dallas) and www.linode.com. I'm getting mid-600 ms ping times.

Not too bad from here (Amsterdam).

Pinging www.linode.com [67.18.92.99] with 32 bytes of data:

Reply from 67.18.92.99: bytes=32 time=118ms TTL=51
Reply from 67.18.92.99: bytes=32 time=118ms TTL=51
Reply from 67.18.92.99: bytes=32 time=119ms TTL=51
Reply from 67.18.92.99: bytes=32 time=118ms TTL=51

Ping statistics for 67.18.92.99:
    Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
    Minimum = 118ms, Maximum = 119ms, Average = 118ms

Traceroute, anyone?

@cirric:

I'm seeing the same problem, for both of my Linodes (which are in Dallas) and www.linode.com. I'm getting mid-600 ms ping times.

Whatever the problem was, it has cleared. I'm now back to ping times < 10 ms.

Yes, I can now get linode.com as fast as other sites, about 70-100ms.

(Regarding traceroute, back when things were slow I tried traceroute both to and from my linode host but didnt get anything after theplanet from the linode side or past our router on my side.)

I hate to be very negative since I really like Linode, but not only is both linode.com and my own site unreasonably slow right now, but this really isn't the first time I've had performance issues, both from my own site as well as linode.com as a whole.

I regularly experience very slow responses from my server, at different times, but usually daily at some point. This whole ordeal at the moment might be another network issue, but other times I believe it's excessive CPU usage or something on my host machine. SSH will have upwards to a second delay at times to my typing, which as you can imagine, kills any productivity.

I do have the smallest Linode account and know my cpu ratio/priority is lowest because of such, and I might be expecting more usage out of it than what I pay for, but when I signed up, it was an excellent service. It's only as of the last few months that it's been regularly crippled like this for me, and what my server runs has changed very little since then. Perhaps Linode simply became more popular in the meantime and is filling up the hosts.

Again I hate to be negative, but this is really just getting to be an aggravating issue, with it being such a regular occurance for me. Hopefully some new hardware in the future to spread the burdon, or the move to this Xen I keep hearing about, will help.

Hi,

@FyberOptic:

I regularly experience very slow responses from my server, at different times, but usually daily at some point.

As do I on my Linode (at HE/Fremont).

Here's my story:

A regular daily sluggishness recently came to my attention when I set up a new cronjob on my local machine. The cronjob would initiate an SSH tunnel (local <-> linode) and then uucp any received mail to my local server.

I soon found that the 09:03 (UTC) job was failing each day (connection timeout) – never a problem at any other HH:03 time.

By process of elimination I concluded that host load that was causing the failure (confirmed as "High" in the Control Panel). I was thinking of filing a support ticket -- to get Linode folk to try and resolve / limit the "other Linode's cron" somehow -- but decided to reschedule my local cronjob to HH:01. That worked around the "09:03" issue so I no longer get a failed connection attempt with the hourly cronjob; but the fact remains that my web/mail servers are extremely sluggish at that time due to host load.

@FyberOptic:

I might be expecting more usage out of it than what I pay for[.] … It's only as of the last few months that it's been regularly crippled like this for me, and what my server runs has changed very little since then.

To partly echo FyberOptic – and to phrase a couple of questions to Chris / Mike / Tom (AKA Linode.com):

Is it really the case that other Linodes will hog the shared resources from time to time and we just need to accept that situation?

Or can something be done? For example, some further product differentiation spread across hosts, like "Linode 100 Pro" and "Linode 100 Burst" -- the former's nodes gets less burstable CPU resources* but the assurance of host/node reachability at all times; the latter for people who amuse themselves with 'emerge -uv gcc glibc' ;-) .

*   I am assuming that CPU resource is the bottleneck; though it may be something else.

Regards,

Cliff</r>

it isn't just network latency..

JobID: 296245 - System Shutdown

Job Entered 12/11/2006 11:58:57 PM Status Success

Host Start Date 12/11/2006 11:59:57 PM Host Finish Date 12/12/2006 12:48:40 AM

Host Duration 48 minutes, 43 seconds Host Message

48 minutes for a shutdown?

nvm found my problem… some search engine critter got lost on my webserver and was abusing mysql/apache

@c1i77:

• I am assuming that CPU resource is the bottleneck; though it may be something else.
  CPU is very unlikely to be the bottleneck, I'd guess IO as I sometimes seem to be able to type in the shell/irc while http access is completely dead.

  Not only is it rather depressing looking at the CPU graph's discontinuities, but I also keep getting messages about the site being down. ;(

@Rakshasa:

Not only is it rather depressing looking at the CPU graph's discontinuities, but I also keep getting messages about the site being down. ;(

Rakshasa: Are you sure that your Linode isn't swap thrashing?

yes, always be sure to check io tokens (cat /proc/io_status). THis was the first thing I did when I noticed extreme slowness on ssh in echoing keystrokes etc. I had a full supply of tokens and was not burning them particularly quickly.

Yes, I've been here long enough to know of those obvious problems.