I second New Relic. I just got it installed, currently on a pair of new servers (in between a migration), and it’s pointed out quite a few things that I never knew before. For example, I found each and every product in a category was pulling over tax information via a complex query… yet the products didn’t even use the tax info. I also discovered performance issues that database caching was hiding from me, because you can view the min and max response time for web transactions. Basically you can see what .php files are performing the slowest, and breakdown that component further into individual queries. Along with this it has your basic slow query and error logs, and scalability reports.
Though, the real question is what to do with the information given. Hopefully you have a web developer that can take the information and fix the issue(s).
I’m not betting with you Patrick !! because there is a high probability that the database is not where it should be!!
There are 2 identical accounts at the server, lets call them A & B
A ends with.biz and B ends with .com.au
This is because the developer set it up mistakenly with a new .biz name and afterwards the owners asked for it to be transferred to .com.au.
BUT the files that are current are in the A account.(I have an altered image there)
Yesterday the host changed the A account name to .com.au and immediately I could not login to the Joomla backend but the shop still could, using an older password. They reported that the orders should be up in the 900’s and were only showing #111.
Also on loading the website an error displayed saying the ssl certificate was self signed and not secure - there had been no problems before that - all this from changing the name!!
From checking it looks as if we are using the folders from one account and the database from the other !!
The host changed the A name back to .biz while we think this through.
New relic sounds exactly what is needed, once I sort out this tangle of accounts and names - but as Jarrod says, “the real question is what to do with the information given”
thanks both
Rob
I went to the New Relic website. I couldn’t get much info on how to install or use it and was reluctant to burn the free 14 day trial without being on top at the start - so I downloaded Microsoft Visual Round Trip Analyser with MS Network Monitor
(For the record I now have the database and the files in the same account and all redirects removed so there should be a clean line between the registrar DNS pointers and the server)
I have performed multiple webpage speed tests since then and there is an amazing variation - TTFB can be anywhere from 3.5 seconds to 20 seconds!!!
here is a Network Monitor display showing what happens.
TTFB here was a ‘quick’ 3.938 seconds
My IP ends with 100 the server ends with 118. Please can anyone explain what is happening here ?
I notice a lot of bad checksums there- if they are nothing to do with TTFB then for the moment I’ll ignore them- there is a lot of optimising to do with the website, but first I must get that TTFB down to under a second
BTW I discovered I have to disable skype, mail etc otherwise there are various pings from them adding confusion to the log.
According to a view source for the website, the first 3 things to load are
[quote]
[/quote]I have disabled the DJ slider which made no difference to ttfb so would I be correct in assuming this delay is BEFORE any part of the website loads? or have I misunderstood the process - I’m unsure what all those ‘payloads’ are before First Byte received.
How does this info compare to New relic? I’d appreciate a basic ‘how to’ on how and where it’s installed, and what it does.
The Visual Round trip analyzer does give a time graph, shows how many ports working and gives ratings and an analysis on the page load (but I expect Patrick that you already know that)
Appreciate any help
Rob
VRTA isn’t really going to provide any more visibility than WebPagetest. Both are measuring on the browser side and the browser doesn’t get any data before TTFB. Yes, the delay is BEFORE any of the website is sent to the browser. The server needs to assemble the HTML it is going to send the browser and that is where the delay is coming from.
You shouldn’t bother with New Relic until you have someone who can act on the data which is probably going to involve changes in how the hosting provider has your account set up. Even if the database and web server are both on the same account, it doesn’t necessarily mean they are on the same server.
I’d start by asking the hosting provider to look at the performance of the server where your mysql is running.
New Relic gets installed on the server and will tell you a lot of information about how the server assembles the pages that are sent to the browser (including each database query, how long it takes and details about the code that is being called). It will also give you a high-level breakdown into the amount of time being used in the database vs application code so you can quickly identify where the bottlenecks are.
Here are the instructions on installing the php agent but I expect it’s mostly going to be Greek to you: http://newrelic.com/docs/php/new-relic-for-php