This thread is a continuation from the original, found here:

http://radified.com/cgi-bin/yabb2/YaBB.pl?num=1203643962

This is not about resource usage specifically, but you may be interested nonetheless (just received from WiredTree):


Quote:

cPanel confirms a root exploit affecting all current installations of Horde (web mail). No fix is currently available. cPanel recommends disabling Horde until they post a patch. We are disabling Horde on all servers. Please do not re-enable it until further notice or it will put your server at severe risk of being compromised. We recommend all users use Squirrelmail until Horde is fixed. If you have previously disabled Squirrelmail, you can re-enable it inside your server's WebHostMgr under Server Configuration >> Tweak Settings. We will post an update in Grove when cPanel has issued a fix and it is safe to re-enable Horde.

Here's a question for Magoo, or one of you othe Linux guru's ..

Are there things my web host (WiredTree) can do to affect the memory/RAM usage on our VPS account?

Seems memory usage is the limiting factor on an account.

Say, for example, we start running close to our limit (384 MB). Are there things you can do as a Linux admin to bump that up tothe limit (and therefore tell the client they need to upgrade to a pricier ho$ting plan)?

Conversely, are there also things they can do to restrict our memory/RAM usage .. and give some of our memory to another account .. and therefore earn more money from our particular server???

If so, would these memory-restricting techniques affect the performance of our account/server.

Or am I being paranoid?

Just seems like when we first opened the account, there were no problems, and memory usage was steady at around 270-280 MB. Now it down around 200-220, and it seems like we've been experiencing performance issues.

What say ye?

Is there anything I can do to ensure we're being properly served? (small pun there) You know the old adage > "You get what you inspect, not what you expect."

http://www.radified.com/gfx3/memory06March.gif

Here's today's load stats .. in case you're interested:

http://www.radified.com/gfx3/load06March.gif

Remember there is a 2-hour difference between here and Chicago.

While I'm at it .. here's the WEEKLY view (of server load):

http://www.radified.com/gfx3/loadWeekly06March.gif

This weekly view shows it going up to only 8 or 9, whereas the daily view shows it hitting 20 or 21. They say the daily view is "most accurate".

Here's a tracert I just ran:

http://www.radified.com/gfx3/tracert06march.gif

Response:


Quote:

It looks like there was a slight load spike that has since passed. Are you still experiencing issues with the server?

The way memory is allocated depends on the virtualization technology.  Some of them statically set aside a certain amount for you.  Some of them just make sure you don't go over your allocation.  Either way, I seem to remember you saying Wired Tree doesn't oversubscribe their host nodes, so you shouldn't have to worry about other sites using your memory.

There's not much they could do to inflate you memory usage, nor do I think they would do it.  It would be easy to detect given the shell access you have and would be a huge risk for them - a potential law suit just to make a few dollars a month.  I don't think either of your hosts have given you false numbers.  Usage dropped on the Lunar Pages server as expected once you moved traffic to the new site, so I think their numbers were real.

It is entirely possible that our intermittent lag is being caused by something other than memory usage.  It could be processor usage, or, most likely it is I/O starvation.  In other words, the hard drive or the front side bus can't keep up with the processor and memory.  From the email Wired Tree sent you, it sounds like another site on your server was causing to much I/O usage and they sent them a warning or some similar situation.  This is all speculation, of course, since I don't have access to either the host node or your site.

Well, the virtualization technology is Virtuozzo 4:

http://www.parallels.com/en/products/virtuozzo/

That much I know.

Memory usage is back up to 276. (MB)

Uh, now it's 292 (max'ing out at 297).

http://www.radified.com/gfx3/memory07March.gif

Memory usage variations probably have a lot to do with cache info.  Apache caches info once it grabs it so that repeated requests for the same page don't cause undue load.  How much memory is used probably depends on how many threads and which ones have been active in the last few days.  I wouldn't worry about it unless you get close to your limit.

Thanks bro.

I'm kinda sensitive to "limits" right now, due to our recent unfortunate experiences at Lunarpages.

Look like our maximum memory usage, so far, is 308 MB:

http://www.radified.com/gfx3/memoryMax308.gif

Here's another statistic they provide, called "Processes" .. tho I'm not sure exactly what this is:

http://www.radified.com/gfx3/processes07March.gif

We usually have somewhere around 100, most of the time. More processes running usually correspond to more memory usage .. tho I haven't analyzed it that closely.

The final metric WiredTree provides is called TCP Sockets. Again, not sure what this is telling me:

http://www.radified.com/gfx3/tcpSockets07March.gif

These numbers remain fairly constant .. to what you see here.

Also noticed we had close to 100 search engines today (mostly Yahoo):

http://www.radified.com/gfx3/searchEngines.gif

A process is an instance of a running program.  For example, when someone requests a page from the website, one of the several running Apache processes gets the file from the disk and sends it to the user.  If Apache detects the current processes are too busy to keep up with user demand, it will start more processes to help out.  Other programs behave similarly.  So, the number of processes is an indication of how busy your CPU is.  Of course, running programs need memory to continue to exist, so each one that is running consumes some memory.  You can limit the maximum number of processes that start if you need to.

A TCP socket is a network connection.  The HTTP protocol used TCP as its underlying network transport protocol.  Each person who wants to communicate with the web site (in order to download a file, for example,) creates a TCP connection to your server.  The connection will stay open for a short time in case the same person wants to use it some more.  After a period of inactivity, the connection will close, freeing the socket.  Network sockets and network ports are similar ideas.  So, the number of open TCP sockets give you an idea of how many computers are communicating with your server via the network.

Thanks for that instructive post. In the case of the graphic posted at the top of this page (identifying more than 60 guests, and nearly 100 search engines, plus a handful of members), wouldn't you expect the number TCP sockets to exceed 30 or 40 .. or even 50?

On a different note, here's a load snapshot taken at 6AM, where I noticed a couple of good-sized peaks (where load hit 6-10).

http://www.radified.com/gfx3/load08March.gif

Here's a new high for our memory usage: 334 MB. Our current web hosting plan provides 256 MB RAM, plus an upgrade for an additional 128 MB. So our MAX limit is 384 MB, at which point we would have to upgrade to a more expensive plan.

http://www.wiredtree.com/managedvps/

So we are 50 MB from our limit.

http://www.radified.com/gfx3/memory08MarchMax334.gif

This daily graphs appears fairly constant and level. Fluctuations based on earlier usage-graphs however, are something I find interesting. I mean, our memory usage has been down below 200 (tho barely) .. while now it's well over 300. That's a sizable fluctuation.

Our memory limit of 384 MB now looms not-so-far away. After dealing with Lunarpages these last 3 months (trying to reduce our usage of server resources) I am sensitive to all server-based limits .. which now seem to focus on our use of memory.

The question now seems to be > at what point do we start to get concerned about memory usage? 50 MB does not seem like very much of a safety buffer. 334 is 87% of our 384 limit. Certainly, 95% usage would require looking into increasing our current memory limit (which would mean paying higher monthly hosting fees).

95% of our 384 MB limit = 365 .. or about 30 MB more .. which makes me reluctant to install Drupal 6 (Content Management System): http://drupal.org/about

.. or at least hesitant. $7.50/month would buy us an additional 64 MB memory. So it's not something we can't afford. (3 less expresso's per month for me.)

Lotsa search engines here this morning (9:30): 105.

http://radified.com/gfx3/searchEngines_105.gif

---

Rad wrote on Mar 8^th, 2008 at 8:35am:

Thanks for that instructive post. In the case of the graphic posted at the top of this page (identifying more than 60 guests, and nearly 100 search engines, plus a handful of members), wouldn't you expect the number TCP sockets to exceed 30 or 40 .. or even 50?

It depends on the timeout value for each.  YaBB might consider someone as being here if they've requested a page in the last 10 minutes, for example.  TCP connections, on the other hand, remain open for as little as a few seconds, so some of the people YaBB considers here may not have an open TCP connection.

Interesting.

Sooo ... would be it be accurate to say (based on the posted metric) that .. at any given instant there are between 30 and 40 people using/accessing the site with an open TCP connection, up to a max of ~50???

If your definition of 'instant' corresponds to how long your server holds tcp connections open (usually anywhere from a few seconds to a few minutes,) then that is an accurate statement.

Here is NEW HIGH for memory use: 366 MB (getting close to our limit of 384):

http://radified.com/gfx3/_memory11March_366.gif

Seems like the sharp spike was quickly reset.

TCP Sockets peak at 134. Interesting. Wonder why.

http://radified.com/gfx3/tcpSockets134.gif

This thread on search engines visiting here is loosely related to this post on TCP Sockets:

http://radified.com/cgi-bin/yabb2/YaBB.pl?num=1207513715

Memory use update (2-day view). Our average seems to have climbed ~20-30 MB.

http://radified.com/gfx3/memory_15apr_1600.gif

The daily view shows us max'ing at 347, which is 90% of our allowable limit of 384-MB:

http://radified.com/gfx3/memory_15apr2008a.gif

Load looks decent (mostly less than 1):

http://radified.com/gfx3/load_15apr2008.gif

We exceeded our allowable memory limit (for the first time), which I mentioned in a separate post, found here:

http://radified.com/cgi-bin/yabb2/YaBB.pl?num=1208353800