0

We recently moved our application from Azure to Google Cloud Platform, in order to better utilize the Firebase platform. We now have a bunch of firebase-functions that communicate with a Redis-instance through a VPC-connector set to max 1000 Mbps. The functionality works, but the performance of Redis-queries are extremely slow at peak hours compared to when we ran it with microservices on Azure.

The Redis-instance we are running on GCP is a standard-instance with 25 Gb of RAM, but only around 13Gb is currently in use. We have about 300.000 keys where all of them are sets or sorted sets that will never expire (ttl=0). The application is constantly adding more items to the various sets.

We don't use Redis as a cache, but rather as a logic-layer to quickly intersect and process data. The intersections can be done between all keys, and because of this, clustering seems like a nightmare, if at all possible.

During peak hours a simple PING in Redis can take over 60 seconds to respond with a PONG. During less active hours, it still can take more than 10 seconds. The CPU-graph looks fine, it sometimes peaks at 100%, but never stays there. Memory fragmentation is at 1.06 according to info memory, which also should be fine according to what I have read. Latency also doesn't seem to be the issue. I have gone through multiple "how to monitor redis performance"-guides, but I am still not closer to figuring out what is wrong.

Does anyone here have any idea at all on what to investigate next?

Below is the output of the info all-command.

# Server
redis_version:4.0.14
redis_git_sha1:00000000
redis_git_dirty:0
redis_build_id:1a41b421362ff783
redis_mode:standalone
os:Linux 4.19.112+ x86_64
arch_bits:64
multiplexing_api:epoll
atomicvar_api:atomic-builtin
gcc_version:6.3.0
process_id:1
run_id:0f04aeebf8790dbf66e8b6b71c95157832e83804
tcp_port:6379
uptime_in_seconds:5372676
uptime_in_days:62
hz:10
lru_clock:15916574
executable:/data/redis-server
config_file:/usr/local/etc/redis/redis.conf

# Clients
connected_clients:386
client_longest_output_list:1
client_biggest_input_buf:0
blocked_clients:0

# Memory
used_memory:13755985544
used_memory_human:12.81G
used_memory_rss:14612512768
used_memory_rss_human:13.61G
used_memory_peak:13821377392
used_memory_peak_human:12.87G
used_memory_peak_perc:99.53%
used_memory_overhead:2717930060
used_memory_startup:3653752
used_memory_dataset:11038055484
used_memory_dataset_perc:80.26%
total_system_memory:35849793536
total_system_memory_human:33.39G
used_memory_lua:3240960
used_memory_lua_human:3.09M
maxmemory:26843545600
maxmemory_human:25.00G
maxmemory_policy:volatile-lru
mem_fragmentation_ratio:1.06
mem_allocator:jemalloc-4.0.3
active_defrag_running:0
lazyfree_pending_objects:0

# Persistence
loading:0
rdb_changes_since_last_save:12652002628
rdb_bgsave_in_progress:0
rdb_last_save_time:1609594635
rdb_last_bgsave_status:ok
rdb_last_bgsave_time_sec:123
rdb_current_bgsave_time_sec:-1
rdb_last_cow_size:717807616
aof_enabled:0
aof_rewrite_in_progress:0
aof_rewrite_scheduled:0
aof_last_rewrite_time_sec:-1
aof_current_rewrite_time_sec:-1
aof_last_bgrewrite_status:ok
aof_last_write_status:ok
aof_last_cow_size:0

# Stats
total_connections_received:11772107
total_commands_processed:44766456481
instantaneous_ops_per_sec:8999
total_net_input_bytes:11667388987
total_net_output_bytes:23685769242
instantaneous_input_kbps:2.23
instantaneous_output_kbps:3.19
rejected_connections:0
sync_full:1
sync_partial_ok:0
sync_partial_err:1
expired_keys:0
expired_stale_perc:0.00
expired_time_cap_reached_count:0
evicted_keys:0
keyspace_hits:7340540
keyspace_misses:265889
pubsub_channels:0
pubsub_patterns:0
latest_fork_usec:302420
migrate_cached_sockets:0
slave_expires_tracked_keys:0
active_defrag_hits:0
active_defrag_misses:0
active_defrag_key_hits:0
active_defrag_key_misses:0

# Replication
role:master
connected_slaves:1
slave0:ip=10.150.95.59,port=6379,state=online,offset=10678179325,lag=3
master_replid:f497652148f651d0c087a9d2bb33fd5f124e18a6
master_replid2:0000000000000000000000000000000000000000
master_repl_offset:10678204720
second_repl_offset:-1
repl_backlog_active:1
repl_backlog_size:2684354560
repl_backlog_first_byte_offset:7993850161
repl_backlog_histlen:2684354560

# CPU
used_cpu_sys:7674.65
used_cpu_user:3521705.50
used_cpu_sys_children:18.38
used_cpu_user_children:120.26

# Commandstats
cmdstat_zadd:calls=44559021486,usec=266567931159,usec_per_call=5.98
cmdstat_get:calls=10595,usec=11200,usec_per_call=1.06
cmdstat_multi:calls=37785398,usec=21279258,usec_per_call=0.56
cmdstat_subscribe:calls=6543,usec=29469,usec_per_call=4.50
cmdstat_config:calls=241673,usec=61201855,usec_per_call=253.24
cmdstat_dbsize:calls=3618,usec=3112,usec_per_call=0.86
cmdstat_sadd:calls=16166,usec=84124,usec_per_call=5.20
cmdstat_psync:calls=1,usec=149,usec_per_call=149.00
cmdstat_bgsave:calls=1,usec=302657,usec_per_call=302657.00
cmdstat_info:calls=2399503,usec=320297314,usec_per_call=133.48
cmdstat_zrangebyscore:calls=2531456,usec=61389137799,usec_per_call=24250.53
cmdstat_zrange:calls=312,usec=20303884,usec_per_call=65076.55
cmdstat_auth:calls=1720895,usec=1254440,usec_per_call=0.73
cmdstat_slaveof:calls=80343,usec=126039,usec_per_call=1.57
cmdstat_lastsave:calls=3,usec=3,usec_per_call=1.00
cmdstat_role:calls=16196665,usec=175267620,usec_per_call=10.82
cmdstat_echo:calls=13088,usec=5588,usec_per_call=0.43
cmdstat_replconf:calls=2087395,usec=2168422,usec_per_call=1.04
cmdstat_ping:calls=13297,usec=5590,usec_per_call=0.42
cmdstat_zrem:calls=258,usec=116323,usec_per_call=450.86
cmdstat_smembers:calls=188,usec=4924728,usec_per_call=26195.36
cmdstat_select:calls=1697357,usec=1228773,usec_per_call=0.72
cmdstat_srem:calls=149,usec=15348,usec_per_call=103.01
cmdstat_exec:calls=142630025,usec=7053257948255,usec_per_call=49451.42
cmdstat_command:calls=3,usec=1355,usec_per_call=451.67
cmdstat_scan:calls=63,usec=1015832,usec_per_call=16124.32

# Cluster
cluster_enabled:0

# Keyspace
db0:keys=312817,expires=0,avg_ttl=0

# oom-prevention
oom_prevention_on:No
current_useconds_with_oom_prevention_on:0
total_useconds_with_oom_prevention_on:0
oom_prevention_threshold:30619509719
oom_prevention_threshold_human:28.52G
used_memory_total:14612512768
used_memory_total_human:13.61G
peak_used_memory_total:15259754496
peak_used_memory_total_human:14.21G
used_memory_aof:0
used_memory_aof_human:0B
used_memory_rdb:0
used_memory_rdb_human:0B

TheQ
  • 101
  • 1
  • Which GCP product are you using Firebase or GCE VM instance? – Serhii Rohoza Jan 05 '21 at 14:02
  • Both :) Queries to Redis are slow from both directions. Firebase-functions connect to Redis through the VPC-connector, but the VM-instances do it directly. When trying to figure this out, I have connected to one of the VM-instances through SSH and executed commands to Redis via redis-cli. – TheQ Jan 07 '21 at 07:27

0 Answers0