Connection Refused

[thiagomorato]

Hey guys I'm getting some connection refused while adding nodes to a layer. The layer is created using the /db/data/ext/SpatialPlugin/graphdb/addSimplePointLayer endpoint with a post method and this data:

{ "layer": "locations", "lat": "lat", "lon": "lon" }

Then to add a point to the layer I use /db/data/ext/SpatialPlugin/graphdb/addNodeToLayer with a post method and with this data (1234 is the node id):

{ "layer": "locations", "node": "http://localhost:7474/db/data/node/1234" }

This work just fine but after doing it a few thousand times I start getting connection refused. It is not related to the data, after 7000 items processed it breaks, if I skip the firsts 6000 it won't break close to the 1000. Does anyone have any idea what might be the issue? I have tried to enable http log but it didn't helped much.

[thiagomorato]

The php exception message is this:

log.ERROR: cURL error 7: Failed to connect to localhost port 7474: Connection refused (see http://curl.haxx.se/libcurl/c/libcurl-errors.html)

From that link it states that error 7 is: Failed to connect() to host or proxy

[thiagomorato]

This might be valuable information:

• Neo4j version: 3.0.4
• Spatial plugin version: 0.19 for neo4j 3.0.3
• Operating system: Mac OSX 10.10
• API/Driver: http
• Steps to reproduce: not sure, it is a bit volatile, maybe it is a matter of configuration. I will try to isolate the issue but if anyone have any tip it would be more than welcome.
• Expected behavior: It shouldn't refuse connections after a few thousand requests.
• Actual behavior: It refuse connections after a few thousand requests.

[thiagomorato]

I have tried switching from the direct http to doing it using the cypher endpoint with procedures, it does a never ending call at item 708, and again if I skip the 600 first itens it won't fail at step 108. I have tried to put some delay between the calls but not joy.

[jexp]

Any chance you can share your data to reproduce? Also any log file showing errors?

Do you run the requests concurrently?

[thiagomorato]

No requests concurrently, well the same strange way it appeared it has gone. If it happens again I will try to isolate it with fake data as I can't share the data I'm working with. The hard thing was exactly this, there was nothing on the logs. Anyway thanks for trying to help : )

[craigtaverner]

Check both the logs/neo4j.log and logs/debug.log for possible errors. Also, if you cannot share your data, perhaps you can create a fake dataset that reproduces this?

[phil20686]

From the regularity and the numbers I bet it triggers a rebalance that is slow enough to time out the http socket. Could you possibly print out the index nodes of the R tree just before it fails?

Sent from my iPhone

On 12 Aug 2016, at 16:06, Craig Taverner [email protected] wrote:

Check both the logs/neo4j.log and logs/debug.log for possible errors. Also, if you cannot share your data, perhaps you can create a fake dataset that reproduces this?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[phil20686]

In fact. Depending on how the http endpoint is implemented, it's possible that you have concurrent modifications. It's a known bug with the current RTree implementation that a that concurrent writes, one of which triggers a rebalance, may result in a concurrent write failure due to the way the locks cascade during a rebalance.

Will the http endpoint accept a second request before the first is complete/tree fully modified?

Sent from my iPhone

On 10 Aug 2016, at 17:03, thiagomorato [email protected] wrote:

Hey guys I'm getting some connection refused while adding nodes to a layer. The layer is created using the /db/data/ext/SpatialPlugin/graphdb/addSimplePointLayer endpoint with a post method and this data:

{ "layer": "locations", "lat": "lat", "lon": "lon" }

Then to add a point to the layer I use /db/data/ext/SpatialPlugin/graphdb/addNodeToLayer with a post method and with this data (1234 is the node id):

{ "layer": "locations", "node": "http://localhost:7474/db/data/node/1234" }

This work just fine but after doing it a few thousand times I start getting connection refused. It is not related to the data, after 7000 items processed it breaks, if I skip the firsts 6000 it won't break close to the 1000. Does anyone have any idea what might be the issue? I have tried to enable http login but it didn't helped much.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.