Open m0n5t3r opened 9 years ago
its very sad to see tokumx just crashed, it should handle the error instead of crashing. i ran into same issue and in my case i found out the reason i.e. my code tries to build an index on very large collection and after 98% it just crashed with same error exception: i tried to increase ulimit , upgrade hardware. but its looks like its the issue with tokumx even in my updated version.
Fri Mar 27 05:01:28.561 Crash reason:
Fri Mar 27 05:01:28.561
Fri Mar 27 05:01:28.561 Got signal: 6 (Aborted)..
i am using 2.0.1. you can see my issue logged here: https://github.com/Tokutek/mongo/issues/1228
we bit the bullet and moved back to mongodb (and upgraded to 3.0 with WiredTiger, which is giving us similar performance and only slightly larger disk usage), 6h of downtime and all... this thing seems dead (they were acquired by Percona some time ago AFAIR, I don't see them pursuing the mongodb thing further)
On 06/04/2015 04:14 PM, Adeel Ahmad wrote:
its very sad to see tokumx is just crashing, it should handle the error instead of crashing. i ran into same issue. in my case i found out the reason, my code tries to build an index on very large collection and after 98% it just crashed with same error exception, Fri Mar 27 05:01:28.561 Crash reason: Fri Mar 27 05:01:28.561
Fri Mar 27 05:01:28.561 Got signal: 6 (Aborted)..
i am using 2.0.1.
you can see my issue logged here: #1228 https://github.com/Tokutek/mongo/issues/1228
— Reply to this email directly or view it on GitHub https://github.com/Tokutek/mongo/issues/1219#issuecomment-108893952.
Thank you very much, i am also trying to test my DB with Mongo 3.0.
i think now is the time to move again from TokuMX to MongoDB :-)
Hello Adeel, Sabin: you are right Percona has acquired Tokutek https://www.percona.com/percona-acquires-tokutek but your statement "I don't see them pursuing the mongodb thing further" is NOT TRUEHere is an official statement from the announcement that we made. WHAT HAPPENS TO TOKUMX?Percona will now offer TokuMX as a fully ACID-compliant drop in alternate replacement to MongoDB®, fully backed by Percona’s renowned support and service offerings. Percona’s TokuMX will provide MongoDB® users with a well known and trusted alternative to existing Support and Consulting services providers. Adeel: It looks like a bug I'm working on a reproducer, it will be nice if you can provide one or your datafrom the data you have provided it looks like your are running out of memory
cacheSize: 5368709120 which smaller that cachetable: size current: 5486165566
your virtual memory is set to ~7G so as workaround you may want to lower cacheSize to like 3G or increase the size of the virtual memory Regards,Abdelhak ErramiSr Support Engineer | Tokutek a Percona companyabdelhak@tokutek.comabdelhak.errami@percona.com
On Thursday, June 4, 2015 9:26 AM, Adeel Ahmad <notifications@github.com> wrote:
Thank you very much, i am also trying to test my DB with Mongo 3.0. i think now is the time to move again from TokuMX to MongoDB :-)— Reply to this email directly or view it on GitHub.
Thank you for replying, i spend last 18 hour and i can reproduce it anywhere and i know its the bug and tokumx should handle that instead of crashing. i can not provide you the data farm but i can help you in providing stats.
i upgraded my hardware , even changed my OS to CentoS 6.6. below are my test result. i tried with 30 GB Physical Memory and with 150GB SWAP and it failed, so i guess its not the virtual memory issue.
1st Test Run for index building
Hard Ware: Amazon Linux , R3.xlarge(4vCPU with 32GB RAM + 150GB SWAP) Disk Space more than 200GB free Space. TokuMX: 2.0.1 Collection catalog_item Size: 108GB Index: db.catalog_item.createIndex( { newItemExtQuoteId : 1 } , { sparse: true }) Ulimit: 64000 Result: TokuMX Crashed.
2nd Test Run for index building
CentOS 6.6, R3.xlarge(4vCPU with 32GB RAM + 150GB SWAP) Disk: More than 500 GB Free Space. TokuMX: 2.0.1 Ulimit: 64000 Collection catalog_item Size: 108GB Index: db.catalog_item.createIndex( { newItemExtQuoteId : 1 } , { sparse: true }) Result: TokuMX Crashed.
@ aerrami:
i tested it further its a bug in tokumx. now i use Server with 120GB RAM and 500GB HDD. i didn't specify cache size which means that line is comment in tokumx.conf. i tried to build the index and it crashed.
what i found is:
i am expecting more from tokumx, crashing is not the option, it should throw some kind of error instead of crashing.
We have a replica set with 2 replicas + arbiter; this morning one of the replicas started crashing (it's gotten to the point where it's immediately after start-up, but initially it did work for a few minutes between crashes)
incoming crash logs:
first crash, the machine was primary at the moment:
last crash, after I saw the notifications, found it dead and attempted to start it: