How would Trains go about sending any missing data to the server in the following scenarios?
Internet connection breaks temporarily while running an experiment
Internet connection breaks and doesn't come back before the experiment ends (any manual way to send all the data that was missed?)
The machine running Trains server resets in the middle of an experiment
Disclaimer: I'm part of the allegro.ai Trains team
Trains will auto retry to send logs, basically forever. The logs/metrics are sent in a background thread so it should not interfere with execution. You can set the backoff parameter, to control the retry frequency, by adjusting the sdk.network.iteration.retry_backoff_factor_sec parameter in your ~/trains.conf file, see example here
The experiment will try to flush all metrics to the backend when the experiment ends, i.e. the process will wait at_exit until all metrics are sent. This means if the connection was dropped, it will retry until it is up again. If the experiment was aborted manually, there is no way to capture/resend those lost metric reports. That said with the new 0.16 version, offline mode was introduced. This way one can run the entire experiment offline, then later report all logs/metrics/artifacts.
The Trains-Server machine is fully stateless (the states themselves are stored in the databases on the machine) this means that from the experiment perspective, the connection was dropped for a few minutes and then it's available again. To your question, if the Trains-Server restarted, it is transparent to all experiments and they continue as usual, no reports will be lost.
Related
I'm familiar with Heroku's policy on dyno restarts (once every 24 hours, if underlying hardware fails, etc).
I've also looked elsewhere on Stackoverflow and found questions like
Controlling Heroku's random dyno restarts.
Our apps are good from a web perspective – sessions are handled via external database and multiple dynos balance the load perfectly. Restarts aren't an issue there. The issue is workers. For example, our worker receives a message and begins processing a job. It's 99% done, awaiting some final asynchronous request to return, and it receives SIGTERM. Before it can even clean up the job, the process is killed. The code can handle local cleanup for a job that needs to be restarted, but external services can't really be part of the transaction.
For example, if a report was built and an email was sent, but some 3rd async operation didn't complete before SIGTERM, I can't really roll back that transaction. For hardware failures or other rare events, it's understandable that a multi-step transaction could get truncated, but with Heroku's policy it seems that I need to assume this will happen at least once a day. Can anyone help me get a better grip on this problem?
I am wondering what would be the best practice for deploying updates to a (MVC) Go web application. Imagine the following scenario :
1) Code and test some changes for my Go Web Application
2) Deploy update without anyone currently using the previous version getting interrupted.
I don't know how to make sure point 2) can be covered - when somebody is sending a request to the server and I rebuild/restart it just in this moment, he gets an error - even if the request just uses a part of the code I did not touch or that is backwards-compatible, or if I just added a new Request-handler.
Maybe I'm missing something trivial or a well-known pattern as I am just in the process of learning go and my previous web applications were ASP.NET- or php-applications where this was no issue as I did not need to restart the webserver on code changes.
It's not just an issue with Go, but in general we can divide the problem into two separate ones:
Making sure current requests do not get terminated and affect user experience.
Making sure there is no down-time in which new requests cannot be handled.
The first one is easier to tackle: You just don't violently kill your server, but tell it to exit, causing a "Drain phase", in which it does not accept new requests and only finishes the currently running requests, and exits. This can be done by listening on signals for example, and entering the app into a special state.
It's not trivial with Go as the default http server doesn't support shutting it down, but you can start a server with a net.Listener, and then keep a reference to it an close it when the time is due.
Now, doing only approach one and then starting the service again will cause new requests not to be accepted while this is going on, and we all know this can take a number of seconds in extreme cases.
So what we need is another instance of the server already running with the new code, the instant the old one is not responding to new requests, right? That can be done in several ways:
Having more than one server, and a load-balancer on top of them, allowing one (or more) server to take the load while we restart another. That's the simplest way, and the way most people do it. If you need N servers to take the load of your users, just keep N+1 and restart one at a time.
Using socket sharing tricks. In Newer Linux kernels, Many processes can listen and accept on the same port. What you do is simply start the new instance and then tell the old one to finish and exit. This way there is no pause. This is done by setting SO_REUSEPORT on the listening socket.
The above can be automated with ready to ship solutions, like Einhorn, that deals with all the details for you, see https://github.com/stripe/einhorn
Another approach is documented in this blog post: http://blog.nella.org/?p=879
I am using Windows Azure SDK 2.2 and have created an Azure cloud service that uses an in-role cache.
I have 2 instances of the service running under normal conditions.
When the services scales (up to 3 instances, or back down to 2 instances), I get lots of DataCacheExceptions. These are often accompanied by Azure db connection failures from the process going in inside the cache. (If I don't find the entry I want in the cache, I get it from the db and put it into the cache. All standard stuff.)
I have implemented retry processes on the cache gets and puts, and use the ReliableSqlConnection object with a retry process for db connection using the Transient Fault Handling application block.
The retry process uses a fixed interval retrying every second for 5 tries.
The failures are typically;
Microsoft.ApplicationServer.Caching.DataCacheException: ErrorCode:SubStatus:There is a temporary failure. Please retry later
Any idea why the scaling might cause these exceptions?
Should I try a less aggressive retry policy?
Any help appreciated.
I have also noticed that I am getting a high percentage (> 70%) cache miss rate and when the system is struggling, there is high cpu utilisation (> 80%).
Well, I haven't been able to find out any reason for the errors I am seeing, but I have 'fixed' the problem, sort of!
When looking at the last few days processing stats, it is clear the high cpu usage corresponds with the cloud service having 'problems'. I have changed the service to use two medium instances instead of two small instances.
This seems to have solved the problem, and the service has been running quite happily, low cpu usage, low memory usage, no exceptions.
So, whilst still not discovering what the source of the problems were, I seem to have overcome them by providing a bigger environment for the service to run in.
--Late news!!! I noticed this morning that from about 06:30, the cpu usage started to climb, along with the time taken for the service to process as it should. Errors started appearing and I had to restart the service at 10:30 to get things back to 'normal'. Also, when restarting the service, the OnRoleRun process threw loads of DataCacheExceptions before it started running again, 45 minutes later.
Now all seems well again, and I will monitor for the next hours/days...
There seems to be no explanation for this, remote desktop to the instances show no exceptions in the event log, other logging is not showing application problems, so I am still stumped.
For being specific, I am using asterisk with a Heartbeat active/pasive cluster. There are 2 nodes in the cluster. Let's suppose Asterisk1 Asterisk2. Eveything is well configured in my cluster. When one of the nodes looses internet connection, asterisk service fails or the Asterisk1 is turned off, the asterisk service and the failover IP migrate to the surviving node (Asterisk2).
The problem is if we actually were processing a call when the Asterisk1 fell down asterisk stops the call and I can redial until asterisk service is up in asterisk2 (5 seconds, not a bad time).
But, my question is: Is there a way to make asterisk work like skype when it looses connection in a call? I mean, not stopping the call and try to reconnect the call, and reconnect it when asterisk service is up in Asterisk2?
There are some commercial systems that support such behavour.
If you want do it on non-comercial system there are 2 way:
1) Force call back to all phones with autoanswer flag. Requerment: Guru in asterisk.
2) Use xen and memory mapping/mirror system to maintain on other node vps with same memory state(same running asterisk). Requirment: guru in XEN. See for example this: http://adrianotto.com/2009/11/remus-project-full-memory-mirroring/
Sorry, both methods require guru knowledge level.
Note, if you do sip via openvpn tunnel, very likly you not loose calls inside tunnel if internet go down for upto 20 sec. That is not exactly what you asked, but can work.
Since there is no accepted answer after almost 2 years I'll provide one: NO. Here's why.
If you failover from one Asterisk server 1 to Asterisk server 2, then Asterisk server 2 has no idea what calls (i.e. endpoint to endpoing) were in progress. (Even if you share a database of called numbers, use asterisk realtime, etc). If asterisk tried to bring up both legs of the call to the same numbers, these might not be the same endpoints of the call.
Another server cannot resume the SIP TCP session of the other server since it closed with the last server.
The MAC source/destination ports may be identical and your firewall will not know you are trying to continue the same session.
etc.....
If you goal is high availability of phone services take a look at the VoIP Info web site. All the rest (network redundancy, disk redundancy, shared block storage devices, router failover protocol, etc) is a distraction...focus instead on early DETECTION of failures across all trunks/routes/devices involved with providing phone service, and then providing the highest degree of recovery without sharing ANY DEVICES. (Too many HA solutions share a disk, channel bank, etc. that create a single point of failure)
Your solution would require a shared database that is updated in realtime on both servers. The database would be managed by an event logger that would keep track of all calls in progress; flagged as LINEUP perhaps. In the event a failure was detected, then all calls that were on the failed server would be flagged as DROPPEDCALL. When your fail-over server spins up and takes over -- using heartbeat monitoring or somesuch -- then the first thing it would do is generate a set of call files of all database records flagged as DROPPPEDCALL. These calls can then be conferenced together.
The hardest part about it is the event monitor, ensuring that you don't miss any RING or HANGUP events, potentially leaving a "ghost" call in the system to be erroneously dialed in a recovery operation.
You likely should also have a mechanism to build your Asterisk config on a "management" machine that then pushes changes out to your farm of call-manager AST boxen. That way any node is replaceable with any other.
What you should likely have is 2 DB servers using replication techniques and Linux High-Availability (LHA) (1). Alternately, DNS round-robin or load-balancing with a "public" IP would do well, too. These machine will likely be light enough load to host your configuration manager as well, with the benefit of getting LHA for "free".
Then, at least N+1 AST Boxen for call handling. N is the number of calls you plan on handling per second divided by 300. The "+1" is your fail-over node. Using node-polling, you can then set up a mechanism where the fail-over node adopts the identity of the failed machine by pulling the correct configuration from the config manager.
If hardware is cheap/free, then 1:1 LHA node redundancy is always an option. However, generally speaking, your failure rate for PC hardware and Asterisk software is fairly lower; 3 or 4 "9s" out of the can. So, really, you're trying to get last bit of distance to the "5th 9".
I hope that gives you some ideas about which way to go. Let me know if you have any questions, and please take the time to "accept" which ever answer does what you need.
(1) http://www.linuxjournal.com/content/ahead-pack-pacemaker-high-availability-stack
Practical Challenge:
I have a LR script that runs against an app being mocked and do not have a logout button (yet).
The test runs fine With stable response time for about 10 minutes, but after that the response time peaks and the server goes into 99% memory usage and transactions start to fail.
I suspect this is due to the script does not terminate the vusers after each run anf it builds up a lot of running sessions against the server wich is not terminated. But I might be wrong.
Anyays I want to programatically close each run after it has competed the business process.
I have red somewhere that web_set_sockets_option ("SHUTDOWN_MODE", "ABRUPT") could be used for this, but I want to be sure that this function actually does what I want and what does 'ABRUPT' means?
Are there better ways of closing sessions? Clicking the close browser during recording does not result in anything being captured in the script.
It's a server issue on session aging. Your server admin for your website can adjust the timeout values where no activity has taken place on a given session. By default most places have this set at 30 minutes. Trim it to what you need rather than taking the default value on the server.
Also, you may have hit a leak situation if resources are constantly accumulated on the server side but never released.
Based on your question I assume you're using the WEB/HTML protocol. I agree that the core issue is that your app's sessions should expire more elegantly and probably sooner. But, in order to get beyond this while testing you can try this. It isn't a guarantee, but it has worked sometimes for me in the past when dealing with similar situations. Try changing your Run-time Settings for the script:
Run-time Settings > Browser > Browser Emulation
Make sure you have the box checked for "Simulate a new user on each iteration". You can also try playing with the other settings here, like clearing the cache each iteration. This could cause a new connection setting with the web page for each iteration depending on the server's session settings. Again, this isn't 100%, but it has worked for me from time to time.
try this:
web_set_sockets_option("CLOSE_KEEPALIVE_CONNECTIONS", "1");