Unfortunately I can not upload my cleanup code but the steps I'm doing
are quite easy. I wrote it in Java using the HDFS API and Curator for
ZooKeeper. Steps are:
- read out the children of /collections in ZK so you know all the
collection names
- read /collections/<collection name>/state.json to get the state
- find the replicas in the state and filter those out that have a
"node_name" matching your locale node (the node name is basically a
combination of your host name and the solr port)
- if the replica data has "dataDir" set then you basically only
need to add "index/write.lock" to it and you have the lock location
- if "dataDir" is not set (not really sure why) then you need to
construct it yourself: <hdfs base path>/<collection name>/<replica
name>/data/index/write.lock
- if the lock file exist delete it
I believe there is a small race condition in case you use replica auto
fail over. So I try to keep the time between checking the state in
ZooKeeper and deleting the lock file as short, like not first determine
all lock file locations and only then delete them but do that while
checking the state.
regards,
Hendrik
On 21.11.2017 19:53, Joe Obernberger wrote:
A clever idea. Normally what we do when we need to do a restart, is
to halt indexing, and then wait about 30 minutes. If we do not wait,
and stop the cluster, the default scripts 180 second timeout is not
enough and we'll have lock files to clean up. We use puppet to start
and stop the nodes, but at this point that is not working well since
we need to start one node at a time. With each one taking hours, this
is a lengthy process! I'd love to see your script!
This new error is now coming up - see screen shot. For some reason
some of the shards have no leader assigned:
http://lovehorsepower.com/SolrClusterErrors.jpg
-Joe
On 11/21/2017 1:34 PM, Hendrik Haddorp wrote:
Hi,
the write.lock issue I see as well when Solr is not been stopped
gracefully. The write.lock files are then left in the HDFS as they do
not get removed automatically when the client disconnects like a
ephemeral node in ZooKeeper. Unfortunately Solr does also not realize
that it should be owning the lock as it is marked in the state stored
in ZooKeeper as the owner and is also not willing to retry, which is
why you need to restart the whole Solr instance after the cleanup. I
added some logic to my Solr start up script which scans the log files
in HDFS and compares that with the state in ZooKeeper and then delete
all lock files that belong to the node that I'm starting.
regards,
Hendrik
On 21.11.2017 14:07, Joe Obernberger wrote:
Hi All - we have a system with 45 physical boxes running solr 6.6.1
using HDFS as the index. The current index size is about 31TBytes.
With 3x replication that takes up 93TBytes of disk. Our main
collection is split across 100 shards with 3 replicas each. The
issue that we're running into is when restarting the solr6 cluster.
The shards go into recovery and start to utilize nearly all of their
network interfaces. If we start too many of the nodes at once, the
shards will go into a recovery, fail, and retry loop and never come
up. The errors are related to HDFS not responding fast enough and
warnings from the DFSClient. If we stop a node when this is
happening, the script will force a stop (180 second timeout) and
upon restart, we have lock files (write.lock) inside of HDFS.
The process at this point is to start one node, find out the lock
files, wait for it to come up completely (hours), stop it, delete
the write.lock files, and restart. Usually this second restart is
faster, but it still can take 20-60 minutes.
The smaller indexes recover much faster (less than 5 minutes).
Should we have not used so many replicas with HDFS? Is there a
better way we should have built the solr6 cluster?
Thank you for any insight!
-Joe
---
This email has been checked for viruses by AVG.
http://www.avg.com
