V20 Prepare two writer processes

Use your current course session. Preparation and course runner · R script ↓

Optional deep dive. Enable only the options named in each block, after preparing its packages or services.

V20 Prepare two writer processes

Goal: Observe coordinated publication.

Prerequisites: V18a; postgres=TRUE; two local R processes with matching packages and TRAINING_PG_CONNECTION.

The current writer lock is catalog-wide and serializes coordinated write sections. This is a manually started two-process experiment, not an automated load test.

training_step(
  "V20",
  {
    writer_config_path <- file.path(training_root, "writer-config.rds")
    saveRDS(pg_config, writer_config_path)
    writer_path <- file.path(training_root, "writer.R")
    writer_code <- c(
      'args <- commandArgs(TRUE)',
      'config <- readRDS(args[1])',
      'writer <- args[2]',
      'data <- data.frame(policy_id=1L, premium=as.numeric(args[3]), cancelled=FALSE)',
      'product <- dataraft.core::dr_product(paste0("training.writer.", writer), data)',
      'print(dataraft.core::dr_publish(product, to=config))'
    )
    writeLines(writer_code, writer_path)
    cat(
      "Terminal A: Rscript ",
      shQuote(writer_path),
      " ",
      shQuote(writer_config_path),
      " A 100
  ",
      sep = ""
    )
    cat(
      "Terminal B: Rscript ",
      shQuote(writer_path),
      " ",
      shQuote(writer_config_path),
      " B 200
  ",
      sep = ""
    )
  },
  needs = c("V18a"),
  enabled = training_options$postgres,
  optional = TRUE
)

Expected result: Two commands to run in separate terminals. Depending on timing, both finish sequentially or one times out on the lock. Tiny deliveries may not visibly overlap.

Task: Can direct SQL bypass the DataRaft lock?