Sticky Sessions: Good or Bad? - session

I have a web application that seems to be having intermittent race conditions when an identical request is sent to two of my load balanced servers. Obviously neither has completed the transaction at this point so both actions on each server is valid.
Would sticky sessions fix this issue? Is the use of sticky sessions frowned upon? Also what might be some other solutions?
I'm hosting right now in Amazon using their load balancer.

If a single request is sent to load balanced set of servers only one of the servers should get the request, typically allocated via round robin. If you are issuing a single request and it hits both your servers something else is wrong.
Otherwise I will assume that you are issuing 2 rapid request and they hit both of your load balanced servers (as round robin would), that your transaction does not complete before the second request hits the server and that you believe sticky sessions would solve this issue.
A sticky session would send all requests in this session to the same server. In your example both requests would now hit the same server and if you did nothing else, the transaction for the first request would not have been committed before the second request started, so you would get the same result i.e sticky sessions alone would not help.
If the transaction were something like placing an order then you could craft your code so that upon successful commit the contents of the cart were deleted.
The first request to complete would delete the cart, the second request would fail and you could message the user that the order had already been placed.
Sticky sessions can make it more complicated to have high availability and scalability. For the former, consider the case where one server goes down - all sessions on that server will also go down and you will have to write code to fail them over to the other server.
For the latter case, assuming your sessions last some interval e.g. 1/2 hour, if you have a N new users come to the site they will initially be evenly divided between both of your servers. If before 1/2 hour all of the users from server 1 leave and another M users come in, then you will have more load on server 2 which has original N/2 user plus new M/2 users while serer 1 only has M/2 users i.e you will have wasted capacity and will need to code to fix.
There are times when sticky sessions may be useful, but unless you have a good reason to use them I would avoid them

Related

Does using a transaction but not actually making any queries have a resource cost?

Ok so one of our team members has suggested that at the beginning of every http request we begin a DB transaction (we are using Entity Framework Core), do the work of the request, and then complete the transaction if the response is 200 Ok, or roll back if it is anything else.
This means we would only commit on successful requests.
That is well and good, when we perform reads and writes to the DB.
However I am wondering does this come at a cost, if we don't actually make any reads or writes to the db?
If you use TransactionScope for this then the transaction is only physically opened on the first database access. The cost for an unused scope is extremely low.
If you use normal EF transactions then an empty transaction will hit the database three times:
BEGIN TRAN
COMMIT
Reset connection for connection pooling
Each of these is extremely low cost. You can test the cost of this by simply running this 100000 times in a loop. It might very well be the case that you don't care about this small cost.
I still would advise against this. In my experience web applications require more flexibility than a 1:1 correspondence of web request and transaction. Also, the rule to use the HTTP status code to decide the transaction will turn out to be inflexible.
Also, you must pick an isolation level (and possibly timeout) for each transaction. At the beginning of an HTTP request it is not known what the right values are. Only the action knows.
I had good experiences with using one EF context per HTTP request and then manually using transactions inside of each action. The overhead in terms of LOC is very small. There is no pressing need to centralize this.
Don't blindly put BEGIN...COMMIT around everything. There are cases where this is just wrong.
What if the web page records the presence of the user, or the loading of the particular page? Having a ROLLBACK destroys that information.
What if there are two actions on the page, and they are independent of each other? That is a ROLLBACK for one is OK, but you want to COMMIT the other?
What if there are no writes on the page? Then there is no need for BEGIN...COMMIT.

Is there a way to keep ajax calls from firing off seemingly sequentially in web2py?

I'm developing an SPA and find myself needing to fire off several (5-10+) ajax calls when loading some sections. With web2py, it seems that many of them are waiting until others are done or near done to get any data returned.
Here's an example of some of Chrome's timeline output
Where green signifies time spent waiting, gray signifies time stalled, transparent signifies time queued, and blue signifies actually receiving the content.
These are all requests that go through web2py controllers, and most just do a simple operation (usually a database query). Anything that accesses a static resource seems to have no trouble being processed quickly.
For the record, I'm using sessions in cookies, since I did read about how file-based sessions force web2py into similar behavior. I'm also calling session.forget() at the top of any controller that doesn't modify the session.
I know that I can and I intend to optimize this by reducing the number of ajax calls, but I find this behavior strange and undesirable regardless. Is there anything else that can be done to improve the situation?
If you are using cookie based sessions, then requests are not serialized. However, note that browsers limit the number of concurrent connections to the same host. Looking at the timeline output, it does look like groups of requests are indeed made concurrently, but Chrome will not make all 21 requests concurrently.
If you can't reduce the number of requests but must make them all concurrently, you could look into domain sharding or configuring your web server to use HTTP/2.
As an aside, in web2py, if you are using file based sessions and want to unlock the session file within a given request in order to prevent serialization of requests, you must use session.forget(response) rather than just session.forget() (the latter prevents the session from being saved even if it has been changed, but it does not immediately unlock the file). In any case, there is no session file to unlock if you are using cookie based sessions.

Why request latency increases when adding more async requests, resulting in ridiculous load time?

I have MVC app that is basically 1 main view and multiple partial views.
I have this tiny script on a min page that loads partials views asynchronously:
<script>
$(function () {
$('#bookmarks').load('Home/Bookmarks');
$('#calendar').load('Home/Calendar');
$('#gmail').load('Home/Gmail');
$('#weather').load('Home/Weather');
$("#news").load("Home/News");
}
</script>
When I comment them all, I get this (very fast loading):
Now I uncomment just the "Bookmarks" request (just reads small JSON file from local drive), I get this (Bookmarks takes 9ms):
Now I uncomment the "Calendar" request (Google Calendar API), I get this (why Bookmarks latency jumps from 9ms to 1.04s if the requests are async?):
Now I uncomment the "Gmail" request (Google Gmail API), I get this (Bookmarks latency jumps again from from 1.04s to 1.53s?):
Now I uncomment the rest ("Gmail", "Weather" and "News" requests), I get these insane increased latencies all over, the Bookmarks request takes now 5s to execute, instead of 9ms - why?):
You can see the increase in latency for each operation - it looks like these ajax requests are not asynchronous at all :( How is that possible, when AJAX is supposed to be async by default?
I am sure I am missing something here, may jQuery load function is not async, but it's on javascript size, and the delay is on server-side. I am now confused.
Update: obviously jQuery call are async, all load functions are executed at the same time. It's easy. The problem is on the server side. After having dome some tests, it's clear that IIS executes these request synchronously, sequentially one after another in the order it received them from the browser. I have done some additional reading on IIS, and by default IIS apppool has only 1 worker process that is being used, that can fire multiple threads for processing all these requests. But for some reason, the requests are being processed sequentially and not in parallel. I haven't found yet why (if a AppPool worker process can start many threads for simultaneous processing) the requests are still executed sequentially, and how to make these requests to be processed in parallel, and if it's even possible. If someone has any idea how to make thing work properly I would really like to hear. Thanks.
Update After some more reading, found that requests are processed sequentially if Session is enabled. The session object is a single threaded object. The session object cannot be shared by two threads simultaneously. Hence when there are two requests for the same session one is queued while the session object is in use by the other. This sucks :( Any suggestions, IIS experts? :)
Solved
Yes, Session State was the problem! I disabled session state in web.config and removed one line that used Session object to avoid runtime error. Now everything works perfect. The true problem was indeed - that Session state is bound to a single thread, therefore on the server side my app behaved like an old STA fart :)
http://msdn.microsoft.com/en-us/library/ms178581.aspx
Concurrent Requests and Session State
Access to ASP.NET session state is exclusive per session, which means that if two different users make concurrent requests, access to each separate session is granted concurrently. However, if two concurrent requests are made for the same session (by using the same SessionID value), the first request gets exclusive access to the session information. The second request executes only after the first request is finished. (The second session can also get access if the exclusive lock on the information is freed because the first request exceeds the lock time-out.) If the EnableSessionState value in the # Page directive is set to ReadOnly, a request for the read-only session information does not result in an exclusive lock on the session data. However, read-only requests for session data might still have to wait for a lock set by a read-write request for session data to clear.
After some more reading, found that requests are processed
sequentially if Session is enabled. The session object is a single
threaded object. The session object cannot be shared by two threads
simultaneously. Hence when there are two requests for the same session
one is queued while the session object is in use by the other. This
sucks :( Any suggestions, IIS experts? :)
Don't use Session. I rarely find cases that justify use of Session, and it's often easy to find storage alternatives that don't have the scalability limitations that you run into with it.

Dealing with concurrency issues when caching for high-traffic sites

I was asked this question in an interview:
For a high traffic website, there is a method (say getItems()) that gets called frequently. To prevent going to the DB each time, the result is cached. However, thousands of users may be trying to access the cache at the same time, and so locking the resource would not be a good idea, because if the cache has expired, the call is made to the DB, and all the users would have to wait for the DB to respond. What would be a good strategy to deal with this situation so that users don't have to wait?
I figure this is a pretty common scenario for most high-traffic sites these days, but I don't have the experience dealing with these problems--I have experience working with millions of records, but not millions of users.
How can I go about learning the basics used by high-traffic sites so that I can be more confident in future interviews? Normally I would start a side project to learn some new technology, but it's not possible to build out a high-traffic site on the side :)
The problem you were asked on the interview is the so-called Cache miss-storm - a scenario in which a lot of users trigger regeneration of the cache, hitting in this way the DB.
To prevent this, first you have to set soft and hard expiration date. Lets say the hard expiration date is 1 day, and the soft 1 hour. The hard is one actually set in the cache server, the soft is in the cache value itself (or in another key in the cache server). The application reads from cache, sees that the soft time has expired, set the soft time 1 hour ahead and hits the database. In this way the next request will see the already updated time and won't trigger the cache update - it will possibly read stale data, but the data itself will be in the process of regeneration.
Next point is: you should have procedure for cache warm-up, e.g. instead of user triggering cache update, a process in your application to pre-populate the new data.
The worst case scenario is e.g. restarting the cache server, when you don't have any data. In this case you should fill cache as fast as possible and there's where a warm-up procedure may play vital role. Even if you don't have a value in the cache, it would be a good strategy to "lock" the cache (mark it as being updated), allow only one query to the database, and handle in the application by requesting the resource again after a given timeout
You could probably be better of using some distributed cache repository, as memcached, or others depending your access pattern.
You could use the Cache implementation of Google's Guava library if you want to store the values inside the application.
From the coding point of view, you would need something like
public V get(K key){
V value = map.get(key);
if (value == null) {
synchronized(mutex){
value = map.get(key);
if (value == null) {
value = db.fetch(key);
map.put(key, value);
}
}
}
return value;
}
where the map is a ConcurrentMap and the mutex is just
private static Object mutex = new Object();
In this way, you will have just one request to the db per missing key.
Hope it helps! (and don't store null's, you could create a tombstone value instead!)
Cache miss-storm or Cache Stampede Effect, is the burst of requests to the backend when cache invalidates.
All high concurrent websites I've dealt with used some kind of caching front-end. Bein Varnish or Nginx, they all have microcaching and stampede effect suppression.
Just google for Nginx micro-caching, or Varnish stampede effect, you'll find plenty of real world examples and solutions for this sort of problem.
All boils down to whether or not you'll allow requests pass through cache to reach backend when it's in Updating or Expired state.
Usually it's possible to actively refresh cache, holding all requests to the updating entry, and then serve them from cache.
But, there is ALWAYS the question "What kind of data are you supposed to be caching or not", because, you see, if it is just plain text article, which get an edit/update, delaying cache update is not as problematic than if your data should be exactly shown on thousands of displays (real-time gaming, financial services, and so on).
So, the correct answer is, microcache, suppression of stampede effect/cache miss storm, and of course, knowing which data to cache when, how and why.
It is worse to consider particular data type for caching only if data consumers are ready for getting stale date (in reasonable bounds).
In such case you could define invalidation/eviction/update policy to keep you data up-to-date (in business meaning).
On update you just replace data item in cache and all new requests will be responsed with new data
Example: Stocks info system. If you do not need real-time price info it is reasonable to keep in cache stock and update it every X mils/secs with expensive remote call.
Do you really need to expire the cache. Can you have an incremental update mechanism using which you can always increment the data periodically so that you do not have to expire your data but keep on refreshing it periodically.
Secondly, if you want to prevent too many users from hiting the db in one go, you can have a locking mechanism in your stored proc (if your db supports it) that prevents too many people hitting the db at the same time. Also, you can have a caching mechanism in your db so that if someone is asking for the exact same data from the db again, you can always return a cached value
Some applications also use a third service layer between the application and the database to protect the database from this scenario. The service layer ensures that you do not have the cache miss storm in the db
The answer is to never expire the Cache and have a background process update cache periodically. This avoids the wait and the cache-miss storms, but then why use cache in this scenario?
If your app will crash with a "Cache miss" scenario, then you need to rethink your app and what is cache verses needed In-Memory data. For me, I would use an In Memory database that gets updated when data is changed or periodically, not a Cache at all and avoid the aforementioned scenario.

Do browsers limit AJAX polling rate? What is the limit?

I just read that some browsers would prevent HTTP polling (I guess by limiting the rate of requests)...
From https://github.com/sstrigler/JSJaC:
Note: As security restrictions of most modern browsers prevent HTTP
Polling from being usable anymore this module is disabled by default
now. If you want to compile it in use 'make polling'.
This could explain some misbehavior of some of my JavaScripts (sometimes requests are just not sent or retried, even if they were actually successful). But I couldn't find further information on details..
Questions
if it's "max. number of requests n per x seconds", what are the usual/default settings for x and n?
Is there any way good resource for this?
Any way to detect if a request has been "delayed" or "rejected" because of a rate limit?
Thanks for your help...
Stefan
Yes, as far as I am aware there is a default pool limit of 10 and a default request timeout of 30 seconds per request, however the timeout and poll limits can be controlled and different browsers implement different limitations!
Check out this Google implementation.
and this is an awesome implementation of catching a timeout error!
You can find the Firefox specifics HERE!
Internet Explorer specifics are controlled from inside the Windows registry.
Also have a look at this question.
Basically, the way you control is not by changing the browser limitations, but by abiding them. So you apply a technique called throttle-ing.
Think of it as creating a FIFO/priority queue of functions. A queue struct that takes xhr requests as members and enforces delay between them is an Xhr Poll. For instance, I am using
Jsonp to get data from a node.js server located on another domain and I am polling of course due to browser limitations. Otherwise, I get zero response back from the server and that is only because of browser limitations.
I am actually doing a console log for every request that's supposed to be sent, but not all of them are being logged. So the browser limits them.
I'll be even more specific with helping you out. I have a page on my website which is supposed to render a view for tens or even hundreds of articles. You go through them using a cool horizontal slider.
The current value of the slider matches the currrent 'page'. Since I am only displaying 5 articles per page and I can't exactly load thousands of articles 'onload' without severe performance implications, I load the articles for the current page. I get them from a MongoDB by sending a cross-domain request to a Python script.
The script is supposed to return an array of five objects with all the details I need to build the DOM elements for a 'page'. However, there are a couple of issues.
First, the slider works extremely fast, as it's more or less a value change. Even if there is drag drop functionality, key down events etc, the actual change takes miliseconds. However, the code of the slider looks something like this:
goog.events.listen(slider, goog.events.EventType.CHANGE, function() {
myProject.Articles.page(slider.getValue());
}
The slider.getValue() method returns an int with the current page number, so basically I have to load from:
currentPage * articlesPerPage to (currentPage * articlesPerPage + 1) - 1
But in order to load, i do something like this:
I have a storage engine(think of it as an array):
I check if the content is not already there
If it is, there is no point to make another request, so go forward with getting the DOM elements from the array with the already created DOM elements in place.
If it isn't, then I need to get it so I need to send that request I was mentioning, which would look something like(without accounting for browser limitations):
JSONP.send({'action':'getMeSomeArticles','start':start,'length': itemsPerPage, function(callback){
// now I just parse the callback quickly to make sure it is consistent
// create DOM elements, and populate the client side storage
// and update the view for the user.
}}
The problem comes from the speed with which you can change that slider. Since every change supposedly triggers a request(same would happen for normal Xhr requests), then you are basically crossing the limitations of all browsers, so without throttle-ing, there would be no 'callback' for most of the requests. 'callback' is the JS code returned by the JSONP request(which is more of a remote script inclusion than anything else).
So what I do is push a request to a priority queue, not POLL, as now I don't need to send multiple simultaneous requests. If the queue is empty, the recently added member is executed and everyone is happy. If it's not, then all non-completed requests in progress are cancelled and only the last one is executed.
Now in my particular case, I do a binary search(0(log n)) to see if the storage engine doesn't have data for the previous requests yet, which tells me if the previous request has been completed or not. If it has, then it's removed from the queue and the current one is processed, otherwise the new one fires. So an and so forth.
Again, for speed consideration and shit browser wanna-bes such as Internet Explorer, I do the above described procedure about 3-4 steps ahead. So I pre-load 20 pages ahead till everything is the client side storage engine. This way, every limitation is successfully dealt with.
The cooldown time is covered by the minimum time it would take to slide through 20 pages and the throttle-ing makes sure there are no more than 1 active requests at any given time(with backwards compatibility going as far as Internet Explorer 5).
The reason why I wrote all this is to give you an example trying to say that you cannot always enforce delay directly from the FIFO structure, as your calls may need to turn into what a user sees, and you don't exactly want to make a user wait 10-15 seconds for a single page to render.
Also, always minimize the polling and the need to poll(simultaneously fired Ajax events, as not all browsers actually do good things with them). For instance, instead of doing something like sending one request to get content and sending another for that content to be tracked as viewed in your app metrics, do as many tasks at server level as you possibly can!
Of course, you probably want to track your errors properly, so your Xhr object from your library of choice implement error handling for ajax and because you are an awesome developer you want to make use of them.
so say you have a try - catch block in place
The scenario is this:
An Ajax call has finished and it's supposed to return a JSON, but the call somehow failed. However, you try to parse the JSON and do whatever you need to do with it.
so
function onAjaxSuccess (ajaxResponse) {
try {
var yourObj = JSON.parse(ajaxRespose);
} catch (err) {
// Now I've actually seen this on a number of occasions, to log that an error occur
// a lot of developers will attempt to send yet another ajax request to log the
// failure of the previous one.
// for these reasons, workers exist.
myProject.worker.message('preferrably a pre-determined error code should go here');
// Then only the worker should again throttle and poll the ajax requests that log the
//specific error.
};
};
While I have seen various implementations that try to fire as many Xhr requests at the same time as they possible can until they encounter browser limitations, then do quite a good job at stalling the ones that haven't fired in wait for the browser 'cooldown', what I can advise you is to think about the following:
How important is speed for your app?
Just how scalable and how intensive the I/O will be?
If the answer to the first one is 'very' and to the latter 'OMFG modern technology', then try to optimize your code and architecture as much as you can so that you never need to send 10 simultaneous Xhr requests. Also, for large scale apps, multi-thread your processes. The JavaScript way to accomplish that is by using workers. Or you could call the ECMA board, tell them to make this a default, and then post it here so that the rest of us JS devs can enjoy native multi-threading in JS:)(how dafuq did they not think about this?!?!)
Stefan, quick answers below:
-if it's "max. number of requests n per x seconds", what are the usual/default settings for x and n?
This sounds more like a server restriction. The browser ones usually sound like:
-"the maximum requests for the same hostname is x"
-"the maximum connections for ANY hostname is y"
-Is there any way good resource for this?
http://www.browserscope.org/?category=network (also hover over table headers to see what is measured)
http://www.stevesouders.com/blog/2008/03/20/roundup-on-parallel-connections
-Any way to detect if a request has been "delayed" or "rejected" because of a rate limit?
You could look at the http headers for "Connection: close" to detect server restrictions but I am not aware of being able in JavaScript to read settings from so many browsers in a consistent, browser-independent way. (For Firefox, you could read this http://support.mozilla.org/en-US/questions/746848)
Hope this quick answer helps?
No, browser does not in any way affect polling. I think what was meant on that page is the same origin policy - you can only access the same host and port as your original page.
Only known limitation to connections themselves is that you usually can only have from two to four simultaneous connections to the same host.
I've written some apps with long poll, some with C++ backend with my own webserver, and one with PHP backend with Apache2.
My long poll timeout is 4..10 s. When something occurs, or 4..10 s passes, my server returns an empty response. Then the client immediatelly starts another AJAX request. I found that some browsers hangs up when I start AJAX call from previous AJAX handler, so I am using setTimeout() with a small value to start the next AJAX request.
When something happens on the client side, which should be sent to server, I use another AJAX request for it, but it's a one-way thing: the server does not send any response, and the client does not process anything. The result of the operation (if any) will be received on the long poll. It requires max. 2 connection to the server, which all browsers supports.
Keep in mind, that if there's 500 client, it means 500 server-side webserver thread, which will move together, occurring load peaks, because when something happens, the server have to report it at the same time for each clients, the clients will process it near same time long, they will start the next long request in the same time, and from then, the timeout will expire also at the same time, and furthcoming ones too. You can trick with rnd timeout, say 4 rnd(0..4), but it's worthless, if anything happens, they will "sync" again, all the request have to be served at the same time, when something reportable happens.
I've tested it thru a router, and it works. I assume, routers respects 4..10 lag, it's around the speed of a slow webapge (far, far away), which no router think, that it should be canceled.
My PHP work is a collaborative spreadsheet, it looks amazing when you hit enter and the stuff is updating simultaneously in several browsers. Have fun!
No limit for no of ajax requests. However it will be on same host & port.
Server can limit no of request from a machine based on its setting.
For example. A server can set so that if there are more than few request from same machine within specified time it will reject request.
After small mistake in javascript code, neverending loop was made witch each step calling 2 ajax requests. In firebug i could see more and more requests until firefox started to slow down, dont response and finally crash.
So, yes, there is a "limit" ;)

Resources