Https://dev.lucee.org/t/increase-in-communications-link-failure-errors/6399

dev support
1

Hi everyone,

I am seeing a lot of the following errors using MySQL datasources in our Prod instance.

"JDBC begin failed: ;JDBC begin failed: ;lucee.runtime.exp.NativeException: JDBC begin failed:
... Caused by: org.hibernate.TransactionException: JDBC begin failed:
... Caused by: com.mysql.cj.jdbc.exceptions.CommunicationsException: The last packet successfully received from the server was 97,851,282 milliseconds ago.
... Caused by: com.mysql.cj.exceptions.CJCommunicationsException: The last packet successfully received from the server was 97,851,282 milliseconds ago.
... Caused by: java.net.SocketException: Connection timed out (Write failed)

I have found this thread;
https://lucee.daemonite.io/t/increase-in-communications-link-failure-errors/6399

Before I went ahead and altered the global timeout, as suggested in that thread - I thought I would shoot off a message and ask if there was a “verified” solution (and reasoning) ?

(In that the thread above has the global timeout change, “I clicked do not use-SSL” and one about bundle names.)

Our PROD datasources ALL have do not use SSL - Staging uses the new driver AND SSL is true

Our prod instance is using v8.0.11 of the driver - which I know is old.
We have upgraded our staging instance to the latest, where I still the same error message on occasion.

Lastly, I read a Java related thread about the same issue and stated to use a datasource “pool”, where the pool checks for a live connection before it is offered to the app. Is this something to investigate?

Thanks in advance.

** MySQL driver : 8.0.11
OS: Amazon Linux 2 (CentOs7)
Java Version: 11
Tomcat Version:8
Lucee Version: 5.3.8.206

2

This is a cloud engineering issue, not a lucee issue.

You have a noisy neighbor. You would do well to provision a transactional vm and place additional caching mechanisms in your application if you have a busy app / site.

QoQ and cachedwithin are 2 of the most powerful features completely underrated by nearly every CF developer.

1 Like
3

@Terry_Whitney! I’m using qoq a lot already :smiley:. When I don’t have too many records in a table, but need extensive filtering for different results (e g. for creating a deeper multilevel navigation drawer). I query all the rows and save it to a components constructor variable. Then I just query the query. Feels way more natural then using fancy reduce() and map() functions with closures.

1 Like
4

I think altering the global timeout on your database will only change the time you need to wait before seeing this error. We run a script every 8 hours with the line below, which solved the issue for us:

<cfset dbpoolclear()>

This is addressed in [LDEV-3139] - Lucee

Simon

5

If I run this command during the middle of the day - will it effect currently running SQL commands / transactions?

Will my active users get an error if they’re doing a DB process?

Thanks!

6

I can’t speak to the details, but we have been using this function for a long time and have never seen an error.
Simon

7

Thanks very much for the quick reply Simon.

Have an awesome day!

8

Does anyone know how database pools work?
WRT to communication failures.

If I have a pool limit of 100 and a timeout of 1 minute;
How does using dbPoolClear() every 8 hours help?
If my connection timeout is 1 minute - or any value less than 8 hours - how will it assist in stopping communication errors?