Aborted (core dumped) in clean new install
supersteves opened this issue · comments
Following the instructions, on both Ubuntu 16.04 and 14.04, 64-bit, installing from latest EventQL binary, using fresh install of latest stable zookeeper.
First run of evqlctl -c /etc/evqld.conf cluster-create
yields just
2017-02-23 16:46:37 evqlctl [INFO] Connecting to zookeeper (192.168.1.216:2181)
with this in zookeeper log
Connection request from old client /192.168.1.216:44268; will be dropped if server is in r-o mode ... caught end of stream exception ... Unexpected Exception: java.nio.channels.CancelledKeyException
Repeat run of same command yields
2017-02-23 16:46:48 evqlctl [INFO] Connecting to zookeeper (192.168.1.216:2181)
Aborting...
<no active exception>
Aborted (core dumped)
with this in zookeeper log
Got user-level KeeperException when processing sessionid:0x15a6bdc33570001 type:setData cxid:0x58af1202 zxid:0xb txntype:-1 reqpath:n/a Error Path:/eventql/mycluster/config Error:KeeperErrorCode = BadVersion for /eventql/mycluster/config
Thanks for the report! I think the problem here might be that the first command succeeded (it currently does not print a good/helpful message on success). The second invocation failed (because the cluster already existed since it was created by the first invocation) but didn't give a good output/error message either.
I'll check if the output/messages are already fixed on master, otherwise we will have to do this. Can you confirm the cluster was successfully created by the first command?
Hi Paul, I did consider that the cluster had worked, and tried to use cluster-list, which also core-dumped.
The only things I can think of that are different to the tutorials:
zookeeper:
dataDir=/home/visokio/zookeeper-3.4.9/data
eventql:
name=stevecluster
zookeeper_hosts=192.168.1.216:2181
and that it's Ubuntu on VirtualBox.
I very much doubt this is relevant.
I think cluster-list aborting might be a separate issue actually. We've been changing the way that cluster joining works (the code is changing towards not connecting directly to zookeeper from the evqlctl binary, but through the evqld process) so the code is not in the best state right now. However, cluster creation and joining servers should work on both the binary release and master, so I'm hoping these are just two unrelated issues.
lol. I didn't realise core dump meant user error! Same for adding a server - do it twice with the same name, second time core dumps. Also confusing was lack of any positive feedback the first time.
Anyhow, thanks - it did indeed work, and I've just completed the tutorial successfully.
Thanks for your help Paul.
Yes, user errors aren't supposed to result in aborts (due to unhandled exceptions) :( Thanks for creating the ticket to fix this - I put it on top of our list.
If there are any more issues/questions please don't hesitate to ping us. Also the master branch does contain a number of bugfixes and performance, however two features are currently not entirely ready so it's not released yet. Still, might be worth to give that a try too if you're experimenting/evaluating EventQL.
Good to know, thank you
@paulasmuth Where do you prefer me to ping you with feedback on my limited success bulk loading 1 billion rows of test data (seg fault on node1 after 130k rows, restarted, but expect I won't get close to 1B if this happens often)?