AEROSPIKE USER SUMMIT 2018 Large Scale NoSQL DB Migration Under Fire Ido Barkan Software Architect AppsFlyer
A E RO S P I K E US E R S UM M I T 2018
Large Scale NoSQL DB Migration Under Fire
Ido Barkan
Software Architect
AppsFlyer
2 AEROSPIKE USER SUMMIT | Proprietary & Confidential | All rights reserved. © 2018 Aerospike Inc
Ag
en
da
Motivation and limitations
What is AppsFlyer?
What have we learned?
War plan
3 AEROSPIKE USER SUMMIT | Proprietary & Confidential | All rights reserved. © 2018 Aerospike Inc
What is AppsFlyer?
~25B events per day
4 AEROSPIKE USER SUMMIT | Proprietary & Confidential | All rights reserved. © 2018 Aerospike Inc
Installs DB- Zoom in
app installs writer
2K write IOPS
app events enrichments 180k read
IOPS
5 AEROSPIKE USER SUMMIT | Proprietary & Confidential | All rights reserved. © 2018 Aerospike Inc
We need to replace CouchBase
● Scale out is out of control(45 r3.4xl instances)
● XDCR can’t keep up● Daily backups take more than a day● Paid support is not good enough
Chosen solution: Aerospike
6 AEROSPIKE USER SUMMIT | Proprietary & Confidential | All rights reserved. © 2018 Aerospike Inc
Starting point
● # of records: ~ 2,000,000,000 (repl. factor 3)● Read intensive: 180K read + 2K write IOPS● # of machines: 45 r3.4xlarge● Cost: ~ 30,000 $ per month● Time frame: a few months.Mission critical
7 AEROSPIKE USER SUMMIT | Proprietary & Confidential | All rights reserved. © 2018 Aerospike Inc
Why is data migration harder than in a relational DB?
Relational DB
NoSql DB
Data types conventional
blob, JSON, specialized bins
Vendors a few 100+
Query Language
SQL N1QL, SQL, JavaScript, HTTP/REST-- creating views/ MapReduce style
Schema Yes Maybe
Migration tools
Yes DIY
8 AEROSPIKE USER SUMMIT | Proprietary & Confidential | All rights reserved. © 2018 Aerospike Inc
Rules of the game
No DowntimeNo data loss
9 AEROSPIKE USER SUMMIT | Proprietary & Confidential | All rights reserved. © 2018 Aerospike Inc
War plan- easy part
● Write new installs/updates to BOTH● Continue reading from the OLD DB Application
10 AEROSPIKE USER SUMMIT | Proprietary & Confidential | All rights reserved. © 2018 Aerospike Inc
Breathing (testing) time
test test test● Sizing RAM/Disk● Expected load● Metrics● Alerting● Logs● Backups
11 AEROSPIKE USER SUMMIT | Proprietary & Confidential | All rights reserved. © 2018 Aerospike Inc
Easy part (Cont.)● Write new installs/updates to NEW DB● Read from BOTH DBs
● A fraction of records (not VIP)
● Do NOT delete yet!● Have metrics on
misses
12 AEROSPIKE USER SUMMIT | Proprietary & Confidential | All rights reserved. © 2018 Aerospike Inc
Breathing time
Test Test Test● Sizing (again)● Expected load
(with reads)● Metrics (again)
13 AEROSPIKE USER SUMMIT | Proprietary & Confidential | All rights reserved. © 2018 Aerospike Inc
Is that it?
If your data is short lived- you are mostly done.
Otherwise...
14 AEROSPIKE USER SUMMIT | Proprietary & Confidential | All rights reserved. © 2018 Aerospike Inc
Dumperview (query)dump keys/records
Kafka as a buffer
Loader-> write(do not overwrite)
values
‘Manual’ partMigrating the long tail
15 AEROSPIKE USER SUMMIT | Proprietary & Confidential | All rights reserved. © 2018 Aerospike Inc
Detour: DB views
function (doc, meta) {emit(meta.expiration, doc.id);}
key (sorting) value
Map function (JS)
● Record itself is NOT a part of the view
16 AEROSPIKE USER SUMMIT | Proprietary & Confidential | All rights reserved. © 2018 Aerospike Inc
‘Manual’ partMigrating the long tail
Dumperview (query)dump keys/records
Kafka as a buffer
Loader-> write(do not overwrite)
values
17 AEROSPIKE USER SUMMIT | Proprietary & Confidential | All rights reserved. © 2018 Aerospike Inc
New Cluster
● # of records: ~ 2,000,000,000 ● 2AZ- repl. factor 2 in each one● IOPS, same and growing● # of machines: 2 x 5 i3.4xlarge● Cost: ~ 10,000 $ per month (⅓)
18 AEROSPIKE USER SUMMIT | Proprietary & Confidential | All rights reserved. © 2018 Aerospike Inc
What have we learned?
1. Plan, Plan, Plan.2. Go easy: test a sample before you start
(1 app / API / Customer).3. Breath (read: test) between stages.4. Do NOT delete until you have to.5. slower -> safer (-> $$$).