Besides serving static files directly, does mechanic expose any commands/tools for adding in caching in nginx? Additionally, are there any gotchas with using nginx's built-in caching with Apostrophe or specific configurations I should use to make sure I'm not borking up core functionality?
I'm the lead architect of Apostrophe at P'unk Avenue.
Mechanic doesn't specifically expose any caching options. You should be able to set up caching via the /etc/nginx/mechanic-overrides folder though, which provides places to insert custom rules at various points in the nginx configuration file that mechanic builds.
As for Apostrophe, there is definitely an issue for administrators editing the site. If you cache the pages, then logging in won't change the appearance of the site to include editing controls. If you make an edit and the edit is cached, you won't see your work. This kind of thing would lead to inconsistent and confusing behavior.
So what I would recommend is using mechanic to set up a separate subdomain of your site just for editing purposes, pointing to the same backend port. The only difference will be that you will not enable caching for it.
This works well but you do have to be careful not to paste any absolute links to the editing subdomain when editing links with the rich text editor.
Then you can cache to your heart's content for the primary domain, as long as you are comfortable with the caching rules you're setting.
Naturally, if you cache the home page for up to a day and then edit the home page, that change will not be immediately reflected on the primary domain.
However, also keep in mind that mechanic is already set up to deliver static files such as media and CSS/JS/font assets directly via nginx, bypassing the backend node process for these. So it's really only necessary to consider caching at the nginx level if you are concerned about the performance of the pages themselves under heavy load.
Speaking of which, you should definitely be running Apostrophe in our multicore configuration, to improve both scalability and reliability:
Running Apostrophe on multiple cores and/or servers
Hope this is helpful!
Related
We're working on an Umbraco site - multiple development machines using a shared development database.
When one developers makes changes in the CMS to content and does a Save and Publish the change is reflected on his machine but not other development machines.
This doesn't seem to make sense as we're all looking at the same database.? We've tried doing an IIS reset to see if it's caching at work but this doesn't seem to make a difference either.
Any ideas what on earth could be going on?
Umbraco does a lot of caching, so it doesn't have to hit the database all the time. Normally, all of the published content is cached in an xml file at App_Data\umbraco.config. You just need to have your developers right click on the root of the content tree in the umbraco backoffice and click "Republish the entire site" to regenerate that xml cache on disk from the xml cache in the database.
You also might need to reindex your examine indexes. You can normally find the "Examine Management" dashboard on the developer section in the backoffice of umbraco. By default, there are three indexes: InternalMember, Internal, and External. Unless you have membership going on in your umbraco site, you can ignore that index. The External index is used mostly for site searches. The Internal index is much more critical. It is used to cache media. I believe it is also used in the backoffice, but I'm not 100% certain. Make sure that the Internal index is regenerated.
Remember that media files are stored in the /media directory by default. That means if developer 'A' uploads a file, the physical file won't show up on developer 'B's machine automatically.
I'll bet you there's some cool ways to set up load balancing to handle a caching for your dev setup. I'm pretty sure there are also ways to store the media in the database, so you don't have to worry about transferring them back and forth.
I wanted to know that, is there some special requirement for a website to make use of CDN ?
i mean is there some special scheme(or atleast considerations) on which your website must be build right from the start to make use of CDN (Content delivery network).
is there anything that can stop a website from making use of CDN, for example the way it references the content files, static file paths or any other thing conceivable.
Thanks
It depends.
You have two kinds of CDN services:
Services like AWS Cloudfront that require you to upload the files in some special place that they read from (eg. AWS S3) - In this case you need have a step in your build process to correctly upload the files and handle the addresses somehow inside your application
Services like Akamai that just need you to change and tweak your DNS records so they will serve the request to your users instead of you - In this case you would have two domains (image.you.com and image2.you.com) and have the image.you.com pointing to Akamai and image2.you.com pointing to the original source of the file. Whenever a user requested an image in Akamai, they would come to you through the "back door", fetch it and starting serving that file always.
If you use the second approach it's really simple to have a CDN supporting your application.
There are a whole bunch of concerns when dealing with CDN solutions.
The first one is that a CDN can't serve a dynamic page - i.e. a page that is unique to every user. Typically, that includes PHP, ASPX, JSP, RubyOnRails etc. - so if you're hoping to support lots of users for a dynamic site, you have to come up with another solution. Some CDN providers support "Edge Side Includes" - this allows you to glue dynamic pages together with cached content on the CDN, but this creates quite a complex application.
Of course, even on a dynamic application, a CDN can still serve static files - images, stylesheets, javascript files, videos etc.
#Tucaz explains the two major options here (actually, Akamai also provides a "filestore" CDN option). If you select the second option - effectively, the CDN becomes a caching reverse proxy in front of your website - it makes sense to tweak the cache headers on your HTTP server, and tell the CDN to honour those. Make sure you set your .ASPX files to not cache!
I currently develop Drupal web sites using its multi-site feature that allows me to have a single code base and support multiple distinct settings per each site.
I set up a dev server and I was quite happy with my arrangement of domains like example.com.local (not that happy because I had to perform a small conversion before entering production, but still quite happy) and the thing used to work well. Too bad I recently started to work at places outside the LAN in which my dev server resides--mostly at clients' places where I need to demo their sites. First of all I set up a dyndns.org account and the server is accessible through the Internet.
Unfortunately the whole domain-based multi-site ungracefully fell down, since I'm now accessing the server via myservername.dyndns.org and Drupal's algorithm takes the domain name into account, so I'm forced to use at least the TLD as part of the directory name (namely sites/local.example.com). So I decided to switch to directory-based multi-site, and now I'm able to access my server from inside the LAN using myservername.local/example.com (having renamed the sites/ subdirectories accordingly). You should easily see why this is suboptimal, since when I browse to myservername.dyndns.org/example.com Drupal looks for sites/org.example.com. I temporarily ended up making a link from sites/org.example.com to sites/local.example.com but again, this does not scale well If and when I'll have to drop dyndns.org for, say, dev.mycorporatesite.com...
Is there any other possibility? I have full access to the server, I can change Apache2's configs, .htaccess and all the stuff.
I would recommend against referencing drupal multisites in folders but instead would set up your server to have a fixed domain name and each site in a subdomain.
So your dev server is at mydevserver.com
and then each site could be
client1.mydevserver.com
client2.mydevserver.com
etc.
If you also at the same time as creating these, you move the files folder from the default to whatever the live site will be i.e.
sites/livesite.com/files
Then when you have to go live, all the references will be correct (if you are drupal 7 this might not be an issue)
I usually place an app_offline.htm in my root directory when I am releasing a website to a production environment. However sometimes if there has been a few big changes to the site, I would like to click around first to make sure it's stable without allowing access to anyone other than me.
As far as I am aware this isn't possible, but I'm hoping someone has a neat solution...
The solution has to include if someone has a deeplink into the site, so using a default.htm/asp page in the root won't do the trick unfortunately.
I agree with the staging environment answer above, but otherwise here's one possible approach: Temporarily block all IP addresses besides your own. This can be achieved through IIS Directory Security configuration, or programmatically in any number of ways
You can redirect all the non-authorized users to an Under Construction page of some sort. Meanwhile, you can happily browse the site from your IP. When the site is vetted, you remove that IP restriction and the site becomes available to the world at large.
It's a difficult thing to achieve. That's why you should have a staging environment where everything should be validated before shipping into production. Then during the deployment process (if it takes long, but it shouldn't) you could use an App_Offline file. This staging environment should be as close as possible to your production environment (in terms of software, patches and configurations installed, not in terms of hardware power of course).
Another quick suggestion that would allow you to control things from the web.config might include a custom module that redirected all requests to a static page except those defined by a filter (i.e. hostname, url sniffing) that could be configured via the web.config.
I have an ASP.NET site that I'm going to have to take down to make some major structural updates to, and I was wondering how I should go about it from the client-side perspective. I have heard of an App_Offline.htm file or something like that, but I've never really gotten that to successfully work. Does anyone know how to do this?
EDIT
My app is running ASP.NET 4.0, for what it is worth.
Rather than messing with the app_offline silliness (among other reasons, you can't continue to see the site internally while performing maintenance), I created an additional "down for maintenance" site in IIS, which is normally stopped. It has the same IP, host headers, etc as the main site, but only has a default.aspx, an images folder and a stylesheet. That file contains the "This site will be down for maintenance until xx:xx PM CST" message.
When I'm ready to perform the update, I stop the main site and start the maintenance site, which then processes any requests it receives and, of course, returns the maintenance message.
When the maintenance is complete, stop the maintenance placeholder site, and restart your main site.
If you're using host headers, you can modify this approach so that the site remains internally accessible over your LAN/WAN while the maintenance site is handling external requests. A simple approach is to remove the host headers for <*>.yourdomain.com from the main site before starting the maintenance site, and ensure that the main site has an additional host header that is internally accessible (added to your local hosts file, for instance). When you start the maintenance site, it'll handle external requests while the primary site will handle requests to the internal-only header.
Alternatively (this seems complex, but saves you the trouble of adding and removing headers), create three sites:
Main site: Configured as in normal operation.
Maintenance site: Has same IP, host headers, etc as main site, but only contains default "down for maintenance" page and any images, css, etc that are required.
Internal test site: Duplicates the configuration of the main site and points to the same folders, but only has host headers,etc for an internal name that is not in the public DNS.
This way, you have only to stop the main site and start the other two in order to funnel external traffic to the "down for maintenance" site, while you can still see and tweak the primary site. This is helpful for that last few minutes of testing/bug fixing that tends to come up during a deployment.
Update
If you don't have access to the server or IIS Manager, you most likely won't be able to use any of that. Assuming that your access is limited to your own folder, your options seem to be either to deploy app_offline.htm to the root of the site (ASP.NET checks for that filename), or to just replace the whole site with a "down for maintenance" app. Perhaps someone else will chime in with alternatives.
The trick for IE is to push over the wire particular count of bytes otherwise IE shows not so friendly 404 error anyway. here is more details: http://weblogs.asp.net/scottgu/archive/2006/04/09/442332.aspx
If you have a good pre-release testing process and careful release procedures you probably won't be down for long.
In that case dropping a file called App_Offline.htm into your site root works fine. IIS will replace your site with it until you remove it. It's a painless way of making sure nothing's updating while you transition.
Mine just has a header with the site logo and a message that we'll be down for maintenance for up to twenty minutes. That's it. It took me about five minutes to write IIRC.
I would definitely recommend this for short sharp down periods of less than half an hour. Anything longer and you're probably looking at a major system change that warrants an approach like David Lively's.