Plumber Manual ⚔

General #

Files: fuser, lsof
Visual: atop, htop
Statistics: dstat (iotop, nethogs and top combined)
Debugging: strace, gcore, bpftrace tools
Lookup/Signal: pgrep, ps, pstree, pkill, killall, kill
Operations: nohup, disown
- nohup : Ignores SIGHUP and then exec the mentioned command
- disown : shell utility that we can use to tell the shell not to send SIGHUP to that process.
Resources
- All my favorite tracing tools

Statistics: df, duf (pretty df), iotop, iostat
File
- inode: stat, file
- content: strings (useful w non-text files), stringsext , filefrag, dd
FS: tune2fs, dumpe2fs, /proc/filesystems
Block device: blkid, findmnt, blockdev (good for getting various sizes), /sys/block/sda/queue/scheduler
Disk: fdisk, smartctl

/proc/meminfo
/proc/buddyinfo
/proc/[pid]/smaps
/proc/[pid]/statm
/proc/[pid]/status

Debugging: traceroute, tracepath, dropreason
Inspection: ss/netstat (what all is running), ip, nethogs (realtime), tcpdump/wireshark/ngrep, iperf3
DNS: drill
Transfer: socat/netcat, rsync
Reconnaissance: nmap
Links
- leandromoreira/linux-network-performance-parameters

Table names
- Should be singular. i.e user instead of users. This also helps with foreign key names(user_id), junction table names (eg. actor_film for actor and film tables) etc.
- But main deal is keeping it consistent, whatever we end up picking. Singular/Plural
- Sometimes may conflict with the database engine keywords, so consider using prefix
Booleans columns
- better represented as nullable Timestamps , Eg. DELETED column instead of bool should be timestamp of timestamp of deletion.
Query building
- Check squirrel for dynamic query building when sqlc is not flexible enough
Migrations
- While dropping tables, drop the ones with fk first when writing migrations
- If developing in golang, if using a migration tool(eg. goose) and sqlc, sqlc’s schema.sql will have duplicate table create statement. It’s fine and OK to have this duplication.
NULL
- If you have a column set to NOT NULL, you probably also want to set a CHECK (column_name <> '') to it aswell, not always but most times.
Primary Keys
- Natural keys / Surrogate keys if applicable
- For identifier column name, if it has to be id, i’d prefer id over <table_name>_id, just cleaner.
- If not using serial keys, use timestamp ordered random ids instead of plain UUID (eg. ULID/UUIDv7)
  - For pg, use pg_idkit / pg_uuidv7
Upserts
- TODO

Don’t think too much and just do the manual conversion of timestamp to protobuf’s format from postgres when needed.
If developing using golang and using sqlc and protobuf, we’d have two different generated structs of the same entity. They might look similar but they’re for different purposes, we want to keep this separated. One is for database and another one is for interchange.