Look into I/O Throttling

[moxious]

From Jake:

version: "2"
​
x-core-common:
  &core-common
  NEO4J_AUTH: neo4j/secret
  NEO4J_dbms_mode: CORE
  NEO4J_ACCEPT_LICENSE_AGREEMENT: "yes"
  NEO4J_causal__clustering_minimum__core__cluster__size__at__formation: 3
  NEO4J_causal__clustering_minimum__core__cluster__size__at__runtime: 3
  NEO4J_causal__clustering_initial__discovery__members: core1:5000,core2:5000,core3:5000
​
services:
  core1:
    container_name: core1
    image: neo4j:4.0-enterprise
    cap_add:
      - NET_ADMIN
    ports:
      - 7474:7474
    blkio_config:
      device_read_bps:
        - path: /dev/mapper/mint--vg-root
          rate: '30mb'
      device_read_iops:
        - path: /dev/mapper/mint--vg-root
          rate: '1920'
      device_write_bps:
        - path: /dev/mapper/mint--vg-root
          rate: '30mb'
      device_write_iops:
        - path: /dev/mapper/mint--vg-root
          rate: '1920'
    environment:
      <<: *core-common
      NEO4J_dbms_connector_bolt_advertised__address: core1:7687
      NEO4J_dbms_connector_http_listen__address: :7474
      NEO4J_causal__clustering_discovery__advertised__address: core1:5000
​
  core2:
    container_name: core2
    image: neo4j:4.0-enterprise
    cap_add:
      - NET_ADMIN
    ports:
      - 7475:7475
    blkio_config:
      device_read_bps:
        - path: /dev/mapper/mint--vg-root
          rate: '30mb'
      device_read_iops:
        - path: /dev/mapper/mint--vg-root
          rate: '1920'
      device_write_bps:
        - path: /dev/mapper/mint--vg-root
          rate: '30mb'
      device_write_iops:
        - path: /dev/mapper/mint--vg-root
          rate: '1920'
    environment:
      <<: *core-common
      NEO4J_dbms_connector_bolt_advertised__address: core2:7687
      NEO4J_dbms_connector_http_listen__address: :7475
      NEO4J_causal__clustering_discovery__advertised__address: core2:5000
​
  core3:
    container_name: core3
    image: neo4j:4.0-enterprise
    cap_add:
      - NET_ADMIN
    ports:
      - 7476:7476
    blkio_config:
      device_read_bps:
        - path: /dev/mapper/mint--vg-root
          rate: '30mb'
      device_read_iops:
        - path: /dev/mapper/mint--vg-root
          rate: '1920'
      device_write_bps:
        - path: /dev/mapper/mint--vg-root
          rate: '30mb'
      device_write_iops:
        - path: /dev/mapper/mint--vg-root
          rate: '1920'
    environment:
      <<: *core-common
      NEO4J_dbms_connector_bolt_advertised__address: core3:7687
      NEO4J_dbms_connector_http_listen__address: :7476
      NEO4J_causal__clustering_discovery__advertised__address: core3:5000

[moxious]

$ sync ; time sh -c "dd if=/dev/zero of=/tmp/testfile bs=100k count=1k  && sync" ; rm /tmp/testfile
1024+0 records in
1024+0 records out
104857600 bytes (105 MB, 100 MiB) copied, 0.326029 s, 322 MB/s
real    0m0.342s
user    0m0.004s
sys    0m0.332s
$ sync ; time sh -c "dd if=/tmp/testfile of=/dev/null bs=4k count=1k  && sync" ; rm /tmp/testfile
1024+0 records in
1024+0 records out
4194304 bytes (4.2 MB, 4.0 MiB) copied, 0.00974453 s, 430 MB/s
real    0m0.024s
user    0m0.004s
sys    0m0.020s

[moxious]

issue is that a lot of cloud providers, when provisioning smaller disks, will give you lower iops and throughput on disks which are small slivers of larger SANs they're using in the background. This can kill neo4j performance. Need some testing & docs on this topic for guidance.