Re: anecdotal uptime and service monitoring

2013-01-29 Thread Jun Rao
In 0.8, we use the metrics package to do the jmx beans and it supports a csv reporter. Thanks, Jun On Tue, Jan 29, 2013 at 5:38 PM, S Ahmed wrote: > Jun, > > Great list. I'm haven't really setup monitoring before, so for starters, > what should I be researching in order to monitor those metr

Re: anecdotal uptime and service monitoring

2013-01-29 Thread S Ahmed
Jun, Great list. I'm haven't really setup monitoring before, so for starters, what should I be researching in order to monitor those metrics, are they exposed via those yammer metrics library that can be exported to a csv file, or are these jmx related items? On Fri, Dec 28, 2012 at 5:47 PM,

Zookeeper caching DNS entries

2013-01-29 Thread Marcos Juarez
We're moving some Zookeeper prod nodes around, and would like to know if the Zookeeper issue below (opened by Neha originally) has been resolved. Ideally we would like to have all the nodes just "see" the new DNS entry, and start talking to the new nodes, instead of having to restart the JVMs.

Re: Payload size exception

2013-01-29 Thread S Ahmed
Ok so it might be an issue somewhere in the pipeline (I'm guessing memory issues?). They are xml files, and that 30-100 was uncompressed. On Tue, Jan 29, 2013 at 12:28 PM, Neha Narkhede wrote: > > At linkedin, what is the largest payload size per message you guys have > in > > production? > > >

Re: Payload size exception

2013-01-29 Thread Jay Kreps
Ack, right you are Neha, my bad. WRT to how to set the maximum there are two considerations: 1. It should be smaller then the fetch size your consumers use 2. Messages are fully instantiated in memory so obscenely large messages (say hundreds of mb) will cause a lot of memory allocation churn/prob

Re: Probably 0.7 bug

2013-01-29 Thread Neha Narkhede
Please can you file a JIRA and attach the full consumer rebalancing log4j file ? On Tue, Jan 29, 2013 at 10:18 AM, Michal Haris wrote: > Hi, just had a strange problem with production kafka cluster of 3 0.7.2 > brokers where a mirror consumer was consuming only 2 of them but when > checking zook

Probably 0.7 bug

2013-01-29 Thread Michal Haris
Hi, just had a strange problem with production kafka cluster of 3 0.7.2 brokers where a mirror consumer was consuming only 2 of them but when checking zookeeper, the missing broker was registered under /brokers/ids but the topic partition was not owned by the consumer. Just looking for a clue while

Re: Payload size exception

2013-01-29 Thread Neha Narkhede
> At linkedin, what is the largest payload size per message you guys have in > production? > Roughly 30K after compression, but that is fairly rare. Most messages are < 500 bytes after compression. Thanks, Neha

Re: Payload size exception

2013-01-29 Thread Xavier Stevens
Not quite in production yet, but we have payloads in the 30KB+ range. I just added a max.message.size to the broker's server.properties. -Xavier On 1/29/13 8:57 AM, S Ahmed wrote: Neha/Jay, At linkedin, what is the largest payload size per message you guys have in production? My app might ha

Re: Jconsole not showing all messages on the broker for a partition

2013-01-29 Thread Jun Rao
All jmx beans are restarted from 0 after restarting the broker. Thanks, Jun On Mon, Jan 28, 2013 at 6:23 PM, Jamie Wang wrote: > Not sure if anyone else ran into this problem in Jconsole. I rely on this > tool to see the number of messages sent by the producer and verifies the > number of mes

Re: Payload size exception

2013-01-29 Thread S Ahmed
Neha/Jay, At linkedin, what is the largest payload size per message you guys have in production? My app might have like 20-100 kilobytes in size and I am hoping to get an idea if others have large messages like this for any production use case. On Tue, Jan 29, 2013 at 11:35 AM, Neha Narkhede wr

Re: Payload size exception

2013-01-29 Thread Neha Narkhede
> In 0.7.x this > setting is controlled by the broker configuration max.message.size. > Actually, in 0.7.x this setting is controlled by max.message.size on the producer. In 0.8, we moved this setting to the broker. Thanks, Neha

Re: Response time degradation

2013-01-29 Thread Neha Narkhede
> if there is a > message in a partition I will always get it in one call? > Yes, if you are talking to the right broker, it will fetch all the available data for all partitions in the multi fetch request. Thanks, Neha

Re: Payload size exception

2013-01-29 Thread Jay Kreps
There is a setting that controls the maximum message size. This is to ensure the messages can be read on the server and by all consumers without running out of memory or exceeding the consumer fetch size. In 0.7.x this setting is controlled by the broker configuration max.message.size. -Jay On T

Re: Response time degradation

2013-01-29 Thread Danny Yeshurun
Will do. Regarding the multi-fetch request - say I give a list of topic-partition, is it guaranteed that all pairs will be "scanned", i.e. if there is a message in a partition I will always get it in one call? Thanks, Danny On Mon, Jan 28, 2013 at 7:54 PM, Neha Narkhede wrote: > Danny, > > Plea