I've recently moved my memcache server behind an Elastic Load Balancer in AWS. I'm also using Flask-Cache with this memcache. If I'm not mistaken (and it's totally possible I am), Flask-Cache opens a connection to memcache and holds it open. It also appears that the ELB terminates these long-standing connections after some period of time (I think it's about 60 minutes). This will result in errors like:
SomeErrors: error 19 from flush_all: (0x4ff96f0) CONNECTION FAILURE, ::rec() returned zero, server has disconnected
If there was some way I could catch these errors and reconnect (or some magic setting to "try to reconnect on connection failure"), that would solve this problem.
FWIW, I'm using pylibmc, but don't see anything obvious (to me) that I could pass.
Any help would be greatly appreciated!
Being disconnected from ELB is very common and also very difficult to debug. Here are a few things that might help:
Debugging Ideas
Attempt to debug the problem in a staging environment with only one
instance connected to ELB.
Make sure you have application logging with time stamps and that if you catch all exceptions in Python (which is generally not a great idea), that you log the exception. It is possible you have a subtle and hidden bug that appears to be something else if you are catching all exceptions.
Simulate the failure (i.e. manually remove "one" instance from ELB), now look at your logs and make sure you see this manifested in your logs. If you can reproduce the same behavior than you can figure out how to fix it.
Look into a web service automated testing tool like https://loader.io/. This can be very helpful to simulate the conditions when the disconnects appear to happen.
Try the same application with a different load balancer, i.e. HAProxy (I would potentially try this last).
Related
I am working on a new Amazon Redshift database that I recently started.
I am experiencing an issue where after I connect to the database, I can run queries without any issue. However, if I spend some time without running anything (like, 5 minutes), when I try running another query or command, ir never finishes.
I am using dBeaver Community 21.2.2 to interact with the connection, and it stays "Executing query" forever. The only way i can get it to work is by cancelling, disconnecting from the redshift, connecting again and then it executes correctly. Until I stop using for some minutes, and then it's happens all over again.
I tought this was a dBeaver issue, as we have a Meabase connected to this same cluster without any issues. But today, I tried manipulating this cluster with R using RJDBC, and the same thing happens: I can run queries, until I stop, and then when I try running something else it never stops, until I disconnect and connect again.
I'm sorry if I wasn't able to explain it clearly, I tried searching for simmilar issues but couldn't.
I suspect that the queries in question are not even being launched on the database. You can check this by reviewing svl_statementtext to see if the query is even being seen. Put a unique comment in the query to help determine if it is actually the query in question.
Since I've seen similar behavior before I'll write up a possible way this can happen. In this case the queries were not being seen by the database or the connection to the database was being dropped mid execution. The cause is network switches and their configurations.
Typical network connections are fairly quick - you ask for a web page and it is given to you. Connection is complete. When you click on a link a new connection is established and also end quickly. These network actions are atomic from a network connection point of view. However, database connections are different. One connection is made and many back and forth transmissions of data happen while the connection is open. No problem and with the right set of network configurations these connections can be open and idle for days.
The problem come in when the operators of the network equipment decide that connections that have no data flowing are "stale" after some fixed amount of time. They do this so that the network equipment can "forget" about these connections and focus on "active" connections. ISPs drop idle connections a lot so that they can handle the load of traffic and connections that flow through their equipment. This doesn't cause any issues for web pages and APIs but database connections get clobbered.
When this happens is look exactly like what you describe. Both sides (client and database) think that the connection is still active but the network equipment has dropped the connection. Nothing gets through but no notification is sent either party. You will likely see corresponding open sessions on the Redshift side for these dropped connections and the database is just waiting for the client to give a command on each of them. An administrator will need to go through and close (terminate) these sessions for them to go away.
Now the thing that doesn't align with experience is the speed at which these connections are being marked as "stale". In my case my ISP was closing connections that were idle for more than 30 min. You seem to be timing out much faster than this. In some cases corporate firewalls will be configured with short idle connection timeouts for routes out of the private network to the internet. So there are cases where the timeouts can be short. The networks at AWS do not have these timeouts so if your connections are completely within AWS then this isn't your answer.
To address this there are a few ways to go. The easy way is to set up a tunnel into AWS with "keep alive" packets sent every 30 sec or so. You will need an ec2 instance at AWS so it isn't cost free. Ssh tunneling is the usual tool for this and there are write-ups online for setting it up.
The hard way (but likely most correct way) is to work with network experts to understand when the timeout is happening and why. If the timeout cannot be changed then it may be possible to configure a different network topology for your use case. Network peering or VPN could address.
In some cases you may be able to not have jdbc or odbc connections at all. You see these protocols are valid but they are old and most networking doesn't work this way anymore which is why they suffer from these issues. Redshift Data API let's you issue SQL to redshift in a single package and check on completion later on. These API calls are each independent connections so there is no possibility of "timing out" between them. The downside is this process is not interactive and therefore not supported by workbenches.
So does this match what you have going on?
I am running a python Django app on an AWS EC2 instance. It uses gunicorn and nginx to serve the app, the EC2 is behind an application load balancer. Occasionally I get 504 error where the entire EC2 instance becomes unreachable for everyone (including via SSH which I use all the time otherwise). I then need to restart everything which takes time.
I can replicate the error by overloading the app (e.g. uploading and processing a very large image), in that case, gunicorn worker times out (I see the timeout message in logs), 504 error appears and the instance becomes unreachable. I set my gunicorn to time out in 5 minutes (300 seconds) but it falls down quicker than that. There is nothing really useful in CloudWatch logs.
I am looking for ways to resolve this for all current and future cases. I.e., I want to have the situation where, if the site gets overloaded, it returns an error message as opposed to becoming completely unreachable for everyone. Is there a way to do that?
There are many things to consider and test here in order to get what is a reason for this, but I think it is OOM(out of memory) mainly because you have to restart even to login in SSH.
Nginx uses "event‑driven" approach to handle requests so a single worker of nginx can handle 1000s of req simultaneously. But Gunicorn on the other hand mostly(by default) uses sync worker which means a request will remain with a worker till it is processed.
When you put a large request your machine tries to process that request until an overflow occurs, mostly it will not get detected by any service running inside a machine. Just try to monitor memory by any monitoring tool in AWS or just SSH and use htop before calling the API.
In most cases with Django/gunicorn the culprit is oom.
Edit:
AFAIK You cannot capture(cache) an oom, the only thing you can do is aftermath i.e after system restart sees/var/logs/syslogs ... As I said monitor in AWS memory monitor(I don't have much experience with AWS).
And regarding the solution,
you first increase the memory of your EC2 until you don't get an error to see how big the problem is.
Then you optimise your application by profiling which part is actually taking this much memory. I haven't used any memory profiling so maybe you can tell me after which is better.
The only thing you can do is optimise your application see common gotchas, best practices, Query optimisations etc.
https://haydenjames.io/how-to-diagnose-oom-errors-on-linux-systems/
https://www.pluralsight.com/blog/tutorials/how-to-profile-memory-usage-in-python
I have a principal database (server_A), mirror database (server_B), and a witness database (server_C). The databases are set up for automatic failover, that is, when server_A goes down or fails over, server_B assumes the role of the new principal database. The database quorum is set up correctly to the best of my knowledge.
I have written an application in c++ to connect to the database and get a value to ensure a true connection. The application detects when a failure occurs on the GetValue call and attempts to reconnect when the error occurs.
The issue is this:
When I have MULTIPLE connections to the database (two threads connected, once connected, it will get a value in a loop), when the failover occurs (stopping sql server on server A so server B will take over as principal), I detect the connection failure and destroy my connection and attempt to reconnect using the same connection string:
"Driver={SQL Native Client};Server=tcp:Server_A;Failover_Partner=tcp:Server_B;Database=SomeDatabase;Uid=SomeUser;Pwd=SomePassword;"
** NOTE **
I have verified that the failover has taken place by monitoring the databases.
Even though, the connection to the database has been properly disposed of, I cannot reconnect to the database until I restart the application, OR if I bring server_A back online (now acting as the mirror database) and then failover server_B (shutting down sql server) making server A the principal database again, the application can reconnect without having to completely close out.
Though I could manipulate the connection string to make server_B the new principal and server_A the new Failover_Partner, this is not an ideal solution as many more connections will be utilized.
Keep in mind, this ONLY happens with multiple connections to the database. If I run the application with only one connection, all is fine and I can reconnect just fine when the failover occurs.
EDIT: If I connect in the beginning with multiple threads, all is fine. When I shutdown SQL Server, and therefore a failover occurs, I can reconnect only when I go through and delete ALL objects and re-instantiate new objects. Also, I am using SQL Native Client 11.0 (ODBC). Thoughts?
A lot of what you're describing is consistent with the issue described in KB 2605597 "Time-out error when a mirrored database connection is created by the .NET Framework data provider for SQLClient."
The KB describes problems when the connection timeout is set to 15 seconds, I have anecdotally heard of similar problems when the connection timeout is set to 0 (which isn't a good idea for other reasons, mentioning just in case).
This hotfix is applied to the application servers. If you want to rule this out as a possible cause, you could test raising the timeout (like it says in the workaround sections of the post) to make sure it's not the issue.
Later thought: The other thing I notice that is unusual is that you're specifying the TCP protocol in the connection string and the failover partner name. It's not clear to me from the documentation that it's supported in the failover partner name. You might want to try removing that and specifying the network attribute instead. (Recommended here.)
I do understand that you believe the issue isn't these things due to the single / multiple connections issue you've tested out.
However, I think you're better off simplifying the connection string so it's as consistent as possible with the published examples and making sure it's not the issues that people have commonly hit with this first. (The retry issue happens when there is latency, which can make it somewhat sporadic.)
Ok I have found the answer.
I had to modify the hosts file because my application did not reside in the same domain as the databases. Therefore when trying to fail over, I could not reach the database with the instance name (which is what the failover partner was cached as). I changed the hosts file to resolve the instance name to the ip address of the machine and it all works now.
To begin, I am running a Wordpress site on an AWS EC2 Ubuntu Micro instance. I have already confirmed that this is NOT an error with Wordpress/mysql.
Seemingly at random the site will go down and I'll get the "Error establishing database connection" message. The server says that it is running just fine, and rebooting usually fixes the issue, however I'd like to figure out the cause and resolve the issue so this can stop happening (it's been the past 2 weeks now that it goes down almost every other day.)
It's not a spike in traffic, or at least Google Analytics hasn't shown the site as having any spikes in traffic (it averages about 300 visits per day.)
What's the cause, and how can this be fixed?
Sounds like you might be running into the throttling that is a limitation on t1.micro. If you use too much CPU cycles you will be throttled.
See http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/concepts_micro_instances.html#available-cpu-resources-during-spikes
The next time this happens I would check some general stats on the health of the instance. You can get a feel for the high-level health of the instance using the 'top' command (http://linuxaria.com/howto/understanding-the-top-command-on-linux?lang=en). Be sure to look for CPU and memory usage. You may find a process (pid) that is consuming a lot of resources and starving your app.
More likely, something within your application (how did you come to the conclusion that this is not a Wordpress/MySQL issue?) is going out of control. Possibly there is a database connection not being released? To see what your app is doing, find the process id (pid) for your app:
ps aux | grep "php"
and get a thread dump for that process: kill -3 to get java thread dump. This will help you see where your application's threads are stuck (if they are).
Typically it's good practice to execute two thread dumps a few seconds apart and compare trends in both. If there is an issue in the application, you should see a lot of threads stuck at the same point.
You might also want to checkout what MySQL is seeing (https://dev.mysql.com/doc/refman/5.1/en/show-processlist.html).
mysql> SHOW FULL PROCESSLIST
Hope this helps, let us know what you find!
I have an Apache2 and Django (mod_wsgi) setup that provides a RESTful API. I have a set of automated tests for this, that executes ~1000 API requests (pure http GET/POST/PUT/DELETE) in sequential order.
The problem is, for every 80 requests or so, I get a strange lag/timeout for exactly 5s or 10s. See timestamp examples here:
Request 1: 2013-08-30T03:49:20.915
Response 1: 2013-08-30T03:49:30.940
Request 2: 2013-08-30T03:50:32.559
Response 2: 2013-08-30T03:50:37.597
I can't figure out why this happens. I have an apache config with KeepAlive Off (recommended setup setting for Django) but otherwise standard install for Ubuntu 12.04 LTS.
I'm running the tests from the same server where the webserver is, I first thought this was some kind of DNS cache thing, but I've added the hostname I'm requesting to /etc/hosts but the problem persists.
The system is idle and have lots of cpu and mem when this lag/timeouts happens.
The lag is not specific to a certain request (URL), it seems kinda random.
Considering that it's always exactly to the millisecond 5s or 10s, it feels like this is some specific setting somewhere causing this.
In case it provides some insight, watch my talk from PyCon US.
http://lanyrd.com/2013/pycon/scdyzk/
The talk deals with things like process churn and startup costs. One thing you shouldn't do is set maximum requests if you don't really need it.
Also consider trying New Relic to help diagnose where the issue is. That will save a lot of guessing if it is a web application of backend service infrastructure issue.
As far as seeing how such monitoring can help, watch another one of my PyCon talks.
http://lanyrd.com/2012/pycon/spcdg/
This was a DNS issue, adding the domainname I used locally to /etc/hosts actually solved the problem. I just hadn't reboot the server for the changes to take effect, thought restarting networking would take care of that, but apparently not.