Marvel creating disk usage imbalance
I now know that Marvel creates a lot of data per day of monitoring - in our case around 1Gb. What I'm just starting to get my head around is the imbalance of disk usage that this caused on my 5 node cluster. I've now removed Marvel and deleted the indexes for now (great tool, but I don't have the disk space to spare on this proof of concept) and my disk usage for the 12 months of rsyslog data has equalised across all the nodes in my cluster. When the Marvel data was sitting there, not only was I using far too much disk space, but I was also seeing significant differences between nodes. At least one node would be using nearly all of the 32Gb, where other nodes would sit at half that or even less. Is there something intrinsically different about Marvel's indexes that makes them prone to such wild differences? Thanks Duncan -- You received this message because you are subscribed to the Google Groups elasticsearch group. To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/7c7d7fb3-a704-4ea5-a74d-efa01f1fa11d%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
Re: Marvel creating disk usage imbalance
Interesting - I thought I'd narrowed it down to Marvel. I had big imbalances with Marvel running, now it all seems flat (although to be fair, the disk usage has dropped to around 5Gb used in a 32Gb partition, so there's large amounts of free space). Same as you though - I could do nothing to rebalance the usage. We've built the cluster so that nodes can be rebuilt and rejoin the cluster if absolutely necessary. Even doing that didn't affect the balance. Will keep an eye on those issues. Looks like there's still a chance of disk imbalance though, if all disks are well below the water mark. Although at least the issue looks to be addressed once disk use pops above the high water mark. Cheers D On Tuesday, 11 November 2014 13:26:01 UTC, Michael Hart wrote: I think it's related to this: https://github.com/elasticsearch/elasticsearch/pull/8270 which I believe was released with 1.4. We see the same thing, with hot spots on some nodes. You can poke the cluster to rebalance itself, which that #8270 fixes permanently, using curl -XPOST localhost:9200/_cluster/reroute. That doesn't always sort it out, and this issue ( https://github.com/elasticsearch/elasticsearch/issues/8149) is our primary issue. AFAIK it's not just Marvel, but any indice can get into this situation. Right now I have a few nodes with 1TB of free disk and others with 400Gb, and Marvel is in another cluster entirely. cheers mike On Tuesday, November 11, 2014 4:15:33 AM UTC-5, Duncan Innes wrote: I now know that Marvel creates a lot of data per day of monitoring - in our case around 1Gb. What I'm just starting to get my head around is the imbalance of disk usage that this caused on my 5 node cluster. I've now removed Marvel and deleted the indexes for now (great tool, but I don't have the disk space to spare on this proof of concept) and my disk usage for the 12 months of rsyslog data has equalised across all the nodes in my cluster. When the Marvel data was sitting there, not only was I using far too much disk space, but I was also seeing significant differences between nodes. At least one node would be using nearly all of the 32Gb, where other nodes would sit at half that or even less. Is there something intrinsically different about Marvel's indexes that makes them prone to such wild differences? Thanks Duncan -- You received this message because you are subscribed to the Google Groups elasticsearch group. To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/3aeaae9e-95c4-4cee-b7dd-1c005e2b965f%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
Re: ES and Java 8. Does it worth the effort ?
We upgraded to Java 8 OpenJDK a few weeks ago - no problems seen so far. D On Friday, 31 October 2014 08:39:39 UTC, Jörg Prante wrote: For JVM support, see http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/setup.html We recommend installing the Java 8 update 20 or later, or Java 7 update 55 or later Jörg On Fri, Oct 31, 2014 at 8:19 AM, Vaidik Kapoor kapoor...@gmail.com javascript: wrote: Would love to hear from Elasticsearch about plans on moving to Java 8. Vaidik Kapoor vaidikkapoor.info On 31 October 2014 03:08, joerg...@gmail.com javascript: joerg...@gmail.com javascript: wrote: Of course is Java 8 worth the effort. Some highlights: - no more permgen OOMs - improved concurrency implementations: http://docs.oracle.com/javase/8/docs/technotes/guides/concurrency/changes8.html - faster hash maps (20%) - lambda expressions are faster than inner classes Some concurrency classes like LongAdder are already included in ES. Java 8 JVM brings also G1 GC to its full extension. G1 GC is not faster than CMS GC, but it scales much better over multicore and reduces stop-the-world pauses to milliseconds. To exploit the full advantage of Java 8, ES would need a large overhaul by rewriting inner classes to lambda style, streams, fork/join pool etc. As long as Lucene does not switch to Java 8, the benefit is only partial. Jörg On Thu, Oct 30, 2014 at 9:01 PM, Georgi Ivanov georgi@gmail.com javascript: wrote: Hi , I wander if i should start using Java 8 with my ES cluster. Are there any benefits using Java 8 ? For example : faster GC , faster Java itself .. anything ES would bebefit from Java 8 .. etc Please share your experience. Georgi -- You received this message because you are subscribed to the Google Groups elasticsearch group. To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearc...@googlegroups.com javascript:. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/9cf82905-63cf-43f2-b14a-de8f21cb4b50%40googlegroups.com https://groups.google.com/d/msgid/elasticsearch/9cf82905-63cf-43f2-b14a-de8f21cb4b50%40googlegroups.com?utm_medium=emailutm_source=footer . For more options, visit https://groups.google.com/d/optout. -- You received this message because you are subscribed to the Google Groups elasticsearch group. To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearc...@googlegroups.com javascript:. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/CAKdsXoGS98nt8%2ByPdm14UHR0pvZVzqVVE%3DtQakQhB_8R6arnNw%40mail.gmail.com https://groups.google.com/d/msgid/elasticsearch/CAKdsXoGS98nt8%2ByPdm14UHR0pvZVzqVVE%3DtQakQhB_8R6arnNw%40mail.gmail.com?utm_medium=emailutm_source=footer . For more options, visit https://groups.google.com/d/optout. -- You received this message because you are subscribed to the Google Groups elasticsearch group. To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearc...@googlegroups.com javascript:. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/CACWtv5mwvxGmww6kQyEtmcq9XjjghAB1LbrGaMJ2gRcF%2BFsJqg%40mail.gmail.com https://groups.google.com/d/msgid/elasticsearch/CACWtv5mwvxGmww6kQyEtmcq9XjjghAB1LbrGaMJ2gRcF%2BFsJqg%40mail.gmail.com?utm_medium=emailutm_source=footer . For more options, visit https://groups.google.com/d/optout. -- You received this message because you are subscribed to the Google Groups elasticsearch group. To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/a04b814e-307c-46ba-aff3-fed79da04f09%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
Refusal to recover after node rebuild
Hi, I've got a proof of concept cluster with 5 nodes. Several months rsyslog data is in there with 2 replicas per index. I then decided to rebuilt 2 nodes simultaneously. No problem. Cluster reallocated as expected and each of the remaining 3 nodes stored all of the indexes and replicas in full. Once the cluster had finished this reallocation, I decided to rebuild another 2 nodes simultaneously (without waiting for the first 2 to come back). This, after all, would leave 1 node storing all of the data. Unfortunately that's where things start to unravel. The initial 2 nodes have come back online and joined the cluster. But not my cluster reports that every shard is unassigned and there doesn't seem to be any process running to reallocate. What I don't understand is that the cluster was fully balanced at 3 nodes and 2 replicas per index. Does taking a node out in this instance cause a problem? My data is still sitting on the node that hasn't been rebuilt, but I can't get it to reallocate onto the other nodes. It's only a proof of concept, so data loss isn't the issue here. It's understanding why this happened and figuring out if I did anything inherently wrong. Cheers Duncan -- You received this message because you are subscribed to the Google Groups elasticsearch group. To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/8a301f07-1634-4b8c-93c7-5a84f45b534e%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
Re: Disk changes forced resync
Not doing any monitoring yet - this is my dev cluster running on 3 workstations. I thought I was quick enough that the rebalance wouldn't have marched ahead and changed much - clearly my admin skills need sharpening! Is there a way to get the cluster to avoid rebalancing when a node is removed from the cluster? I wouldn't want a cluster rebalance starting just because I'm patching the OS and need a reboot. Thanks Duncan -- You received this message because you are subscribed to the Google Groups elasticsearch group. To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/8eed4547-1ca7-4fc7-b291-529653618f94%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
Disk changes forced resync
Apologies if this is a silly question. I recently changed the disk layout on one of my ES nodes to put /var/lib/elasticsearch on it's own disk partition. Around 100Gb data was set to one side, new disk created, then rsync'd to it's new partition. As far as I can be certain, everything was the same before and after the operation - just that the data now sat on it's own partition. Ownership, permissions, ACL's, SELinux contexts were all synced. When I started elasticsearch back up again, however, the 100Gb in the partition vanished and the node started to rebuild itself from scratch, copying data in from all the other nodes. The time between stopping and restarting elasticsearch was around 15 minutes. I expected the data that I'd put back in place to be used first as most of it is historical and won't have changed. Did I do something wrong in my procedure, or am I just expecting the wrong thing. I'm hoping that doing disk maintenance like this in a production system doesn't trigger such a rebuild as my prod systems will have significantly more data. Many thanks Duncan RHEL 6.5 x86_64 java-1.7.0-oracle-1.7.0.51-1jpp.1.el6_5.x86_64 elasticsearch-1.1.0-1.noarch -- You received this message because you are subscribed to the Google Groups elasticsearch group. To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/e0a204a7-318b-447b-8fcb-799b5a3b%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.