@thisisstephenbetts I think the problem here is simply that the server is not ready. We are using less and less neo4j-rake_tasks in favor on boltkit. In any case you need to implement a waiting loop for the server like you can see e.g. activegraph/travis.yml https://github.com/neo4jrb/activegraph/blob/ca2fade3f37b2fe00be6f26a3aa0e63e06734fc4/.travis.yml#L14
Thanks @klobuczek — you might be right, but I don't think so. If I remove that sleep statement I get a
Neo4j::Driver::Exceptions::ServiceUnavailableException error, not the session expired one. Also I have increased that sleep to 5 minutes for testing, with no difference. Finally, you can see that going from a failed query to a successful one is virtually instantaneous — there is no additional sleep statement. If the problem was that the sleep wasn't long enough I'd expect the 2nd query to fail sometimes, but I've never seen that.
I will explore boltkit — thanks for the tip
I modified the code above to be (not an exact copy, but close enough) to
Rake::Task['neo4j:start'].invoke(environment)
sleep 1 while `curl localhost:7575 > /dev/null 2>&1 ; echo $?`.chomp != "0"
tried_before = false
begin
ActiveGraph::Base.query("Match (n:ModelVersion) return n.version").entries
rescue Neo4j::Driver::Exceptions::SessionExpiredException
puts "retrying"
raise if tried_before
tried_before = true
retry
endand I see the "retrying" output, exactly once and everything is fine after that.
As far as I'm concerned this is an okay workaround for now, so I'm mostly just posting here for your information, not to obtain an immediate fix. Thanks!
@klobuczek to me all the evidence seems to say it's not a timing issue — if nothing else I believe you'd get a different exception. If I had to guess I'd say whatever is supposed to detect expired sessions has a small bug, but — not knowing the underlying code — that's not based on much.
Here you go — this is a compact script to demonstrate the problem (you may need to modify the http URL to test for availability, but apart from that it should be portable). Uses Neo4j 4.0.6, Ruby 2.6.3, Rails 5.2.3, and ActiveGraph 10.0.1
require 'active_graph/railtie'
Rake::Task['neo4j:start'].invoke('test')
Rake::Task['neo4j:start'].reenable
Timeout.timeout(120) do
sleep 1 while `curl localhost:7575 > /dev/null 2>&1 ; echo $?`.chomp != "0"
end
ActiveGraph::Base.query("Match (n:ModelVersion) return n.version").entries
Rake::Task['neo4j:stop'].invoke('test')
Rake::Task['neo4j:start'].invoke('test')
Timeout.timeout(120) do
sleep 1 while `curl localhost:7575 > /dev/null 2>&1 ; echo $?`.chomp != "0"
end
tried_before = false
begin
ActiveGraph::Base.query("Match (n:ModelVersion) return n.version").entries
rescue Neo4j::Driver::Exceptions::SessionExpiredException
puts 'Got SessionExpiredException'
raise if tried_before
tried_before = true
retry
end
puts ActiveGraph::Base.query("Match (n:ModelVersion) return n.version").entries.length
@thisisstephenbetts when you start or restart the neo4j server the connections in the driver pool become stale. The easiest workaround is to
ActiveGraph::Base.driver = nil after start. You will still need to wait for the server to start. Open question: why is the driver initialized before server start? Are you making any queries before? Ideally the stale connection should be purged and new one created. This is working in the java based drivers 1.7 and 4.1, but seems not to be on the seabolt based driver 1.7.
$ jruby -S rails generate scaffold User name:string email:string
2020-11-13T11:37:59.585-05:00 [main] WARN FilenoUtil : Native subprocess control requires open access to the JDK IO subsystem
Pass '--add-opens java.base/sun.nio.ch=org.jruby.dist --add-opens java.base/java.io=org.jruby.dist' to enable.
2020-11-13T11:38:03.555-05:00 [main] WARN FilenoUtil : Native subprocess control requires open access to the JDK IO subsystem
Pass '--add-opens java.base/sun.nio.ch=org.jruby.dist --add-opens java.base/java.io=org.jruby.dist' to enable.
invoke active_graph
identical app/models/user.rb
ArgumentError: Can not transliterate strings with Windows-1252 encoding
transliterate at D:/jruby-9.2.13.0/lib/ruby/gems/shared/gems/activesupport-6.0.3.4/lib/active_support/inflector/transliterate.rb:67
parameterize at D:/jruby-9.2.13.0/lib/ruby/gems/shared/gems/activesupport-6.0.3.4/lib/active_support/inflector/transliterate.rb:123
parameterize at D:/jruby-9.2.13.0/lib/ruby/gems/shared/gems/activesupport-6.0.3.4/lib/active_support/core_ext/string/inflections.rb:196
base_migration_file_name at D:/jruby-9.2.13.0/lib/ruby/gems/shared/gems/activegraph-10.0.1/lib/rails/generators/active_graph_generator.rb:15
migration_file_name at D:/jruby-9.2.13.0/lib/ruby/gems/shared/gems/activegraph-10.0.1/lib/rails/generators/active_graph_generator.rb:19
migration_template at D:/jruby-9.2.13.0/lib/ruby/gems/shared/gems/activegraph-10.0.1/lib/rails/generators/active_graph_generator.rb:40
create_model_file at D:/jruby-9.2.13.0/lib/ruby/gems/shared/gems/activegraph-10.0.1/lib/rails/generators/active_graph/model/model_generator.rb:21
run at D:/jruby-9.2.13.0/lib/ruby/gems/shared/gems/thor-1.0.1/lib/thor/command.rb:27
invoke_command at D:/jruby-9.2.13.0/lib/ruby/gems/shared/gems/thor-1.0.1/lib/thor/invocation.rb:127
invoke_all at D:/jruby-9.2.13.0/lib/ruby/gems/shared/gems/thor-1.0.1/lib/thor/invocation.rb:134
each at org/jruby/RubyHash.java:1415
map at org/jruby/RubyEnumerable.java:886
invoke_all at D:/jruby-9.2.13.0/lib/ruby/gem
Has anyone gotten libseabolt17 to install to a debian container? Alternatively, has anyone gotten it to install to a CentOS 8 container? Those are the two we've set up our app to run upon and I'm really not excited about trying to get "yet another" container platform working...
Hello @klobuczek . 3.3.2020 you said that "the 1.7 drivers even the official ones do not support multiple databases, the driver can only access the default database. Java driver 4.0 can do it, but we are not at that level yet, seabolt does not support it. I have an open neo4j support ticket suggesting simple server side fix to that"
Could you please provide a link to this server side fix description because we can't find it? Or could you say something more about it? Thank you!
CC @ekampp
Could you please provide a link to this server side fix description because we can't find it? Or could you say something more about it? Thank you!
CC @ekampp
@jacob-ablowitz check that :
RUN apt-get install -y wget
RUN wget https://github.com/neo4j-drivers/seabolt/releases/download/v1.7.4/seabolt-1.7.4-Linux-ubuntu-18.04.deb
RUN dpkg -i seabolt-1.7.4-Linux-ubuntu-18.04.deb
RUN rm seabolt-1.7.4-Linux-ubuntu-18.04.deb
@kapone89 I asked neo4j to allow the addtion of query parameter to the server url e.g. neo4j://neo01.graph.example.com?database=sales like they already do with routing context neo4j://neo01.graph.example.com?policy=europe (https://neo4j.com/docs/driver-manual/current/client-applications/#driver-routing-context) but they didn’t want to do it.
This was an internal support ticket.
@kapone89 we include something like that in our controllers not for the purpose of switching databases but to ensure causal cluster replication level. Every application using causal cluster should have something like that not neccessarily wrapping entire action but somewhere inside the action. You could have multiple such session blocks addressing different databases.
included do
around_action :wrap_in_session_or_transaction
end
def wrap_in_session_or_transaction
ActiveGraph::Base.session(bookmarks: transaction_bookmarks) do
wrap_in_transaction? ? ActiveGraph::Base.send(transaction_method) { yield } : yield
end
@last_bookmark = serialize_bk(ActiveGraph::Base.last_bookmark)
auth_token_headers
endtransaction_method is read_transaction or write_transaction. You need to specify this correctly to get the full benefit of causal cluster.
Hi. We have an array field in our nodes and I want to search by this field.
If I do Node.where(ar: [1, 2, 3]) the query is ... WHERE n.ar IN $ar..., how can I use the equality operator = instead of IN?
I tried making a custom type with a custom converter where to_db returned an array, but even then the query used IN but not =.
Now you are crossing the line. Sounds like you have semantic composite keys. Rails with ActiveRecord took the approach of having single id field if necessary in addtion to the semantic keys. ActiveGraph follows the same approach. It should not be attempted to work around this by encoding the composite keys into single id field. Another aspect to consider is if neo4j would properly index such an array property, which I don’t know from the top of my head.
Anyone else using Neo4j Aura with ActiveGraph? I got an email today saying:
"It seem you are using a Ruby driver (we believe via http://neo4jrb.io/ ) to connect to Aura and as you know unfortunately this driver is not part of the official Neo4j supported drivers certified to work well with Aura.
Note that this driver https://github.com/neo4jrb/activegraph#neo4j-version-support states that support for 4.1 and above are not supported and since Aura is running 4.2+ this is indeed an issue."
Anyone else got this?
And are you using seabolt driver or jruby driver?
The only driver that supports 4.2 is https://rubygems.org/gems/neo4j-java-driver/versions/4.2.0.beta.0-java and https://rubygems.org/gems/activegraph/versions/11.0.0.beta.4-java
I've recently had an issue where an association defined in an ActiveRel class and validated on the from_class like"validates :assoc, presence: true" fails, but when I remove the ActiveRel and replace with rel type defined on the from class the validation works. This is with version 9.6.2. Any similar experiences out there?
