Why your origin logs are no longer the whole story
Before a CDN, your web server saw every request and wrote it to its access log. After a CDN, the edge answers cached requests itself, often most of them, and your server sees only what it had to serve: cache misses, uncacheable pages, API calls. Those requests arrive from edge addresses too, unless you restore the visitor address from a header, as the real visitor IP guide shows.
The complete record of who asked for what now lives at the edge. Log delivery sends that record to you: every request, hit or miss, with the visitor's address, the timing and the cache status.
What log delivery gives you
Answers to specific questions. What happened to this one request? Which addresses downloaded this file? When did the 5xx errors start, and on which URLs? Reports summarise; logs answer.
Security and investigations. A SIEM correlates edge logs with your application, firewall and login logs, so you can follow an attack, a scraper or a leaked link across days.
Your own analytics. Load the lines into a data warehouse and build the reports your business needs, by country, network, device or content type.
A record you control. The files are yours: keep them as long as your policies require, in the system you already use.
Anatomy of a log line
Log delivery services usually write JSON Lines: one complete JSON object per line, one line per request, gzip-compressed. The format streams, splits and appends cleanly, and every log tool reads it. On cdn.com.tr a line carries these fields:
When and where: ts (the time, ISO 8601 with offset), edge and edge_port (the edge address and port the visitor connected to).
Who: client_ip and client_port (the visitor's address and source port), country, asn (the visitor's network), user_agent and referer.
What: method, scheme, host, uri and protocol.
How it went: status, bytes (body bytes sent), cache (HIT, MISS and the other cache states), request_time (seconds the edge spent on the request) and origin_time (seconds your origin took, - on a cache hit).
The source port matters more than it looks. In mobile networks and behind carrier-grade NAT, many visitors share one public address; only the address, the port and the time together point to one connection.
One request, one line (wrapped here for reading)
{"ts":"2026-11-20T11:03:07+00:00","edge":"203.0.113.10","edge_port":"443",
"client_ip":"198.51.100.7","client_port":"53114","method":"GET","scheme":"https",
"host":"www.example.com","uri":"/products/42","protocol":"HTTP/2.0",
"status":200,"bytes":18342,"cache":"HIT","request_time":0.004,"origin_time":"-",
"referer":"https://www.example.com/","user_agent":"Mozilla/5.0 ...",
"country":"TR","asn":64500}
Logpush on cdn.com.tr
Access logs are a switch on your CDN account's Access logs page, and one switch covers every domain the account serves through the edge. With delivery on, each edge collects your requests and uploads a gzip-compressed JSON Lines file every five minutes; the first files appear within about ten minutes.
Files are grouped by day (UTC), and each covers five minutes of one edge. Download single files or a whole day from the panel, or work from the command line: cdnctl logs pull fetches a day and, run again, only the files that are new; cdnctl logs grep downloads just the files in a time window and filters them on your computer by IP, status code, path and more. Download links point only inside your account's own folder and expire after 15 minutes. Owners and editors switch delivery on and off; every team role can list and download.
Files are kept for 30 days and then deleted automatically. For a longer history, pull them into your own archive on a schedule, as shown below. The access logs guide has every step.
IP masking
Access logs contain personal data: a visitor's IP address can identify a person, and as the site owner you are the data controller for these logs. Decide before switching delivery on whether you need full addresses.
With IP masking on, the edge writes only the network part of each address: an IPv4 address keeps its first three octets (198.51.100.0, the /24) and an IPv6 address its first three groups (the /48). That is enough for country, network and abuse statistics and keeps individual visitors out of your files, while the source port still tells connections apart. When you need full addresses, for investigations or a record-keeping duty, leave masking off and restrict who can read the files.
Getting logs into your SIEM or archive
Treat the 30-day window as a delivery buffer and your own storage as the record. A scheduled job that pulls new files is all it takes, and because cdnctl logs pull downloads only the files you do not have yet, running it every hour costs almost nothing.
From there, send the files wherever they belong: to Elasticsearch or OpenSearch with Filebeat or Logstash, into BigQuery or ClickHouse as JSON, to your SIEM's file input, or to object storage with a lifecycle rule for long-term retention. For a tamper-evident archive, record a checksum of each file, or sign it, as it arrives.
Pull new files every hour, then query them
# crontab: today's new files every hour, yesterday's last files after midnight (UTC days)
ACCOUNT_UUID=your-account-uuid
0 * * * * cdnctl logs pull --account $ACCOUNT_UUID --out /var/log/cdn
20 0 * * * cdnctl logs pull --account $ACCOUNT_UUID --day $(date -u -d yesterday +\%F) --out /var/log/cdn
# 5xx errors with time, host and URL
zcat /var/log/cdn/*.jsonl.gz | jq -r 'select(.status >= 500) | [.ts, .host, .uri, .status] | @tsv'
# one network in a time window, straight from the CLI
cdnctl logs grep --account $ACCOUNT_UUID --from "2026-11-20 14:00" --to "2026-11-20 15:00" --ip 198.51.100.0/24 --format table
Who reads edge logs
Edge requests in the SIEM next to firewall and login events, so an attack can be followed from the first probe.
The exact status, cache state and timing of the request a customer complains about, in seconds.
A complete request history in your own storage, kept for as long as your policies require.
How to set it up
Step-by-step guides in the help centre:
Logpush FAQ
What is the difference between Logpush and analytics reports?
Reports aggregate requests, bandwidth and cache ratio over time. Logs keep every single request with its own fields, so you can answer questions no report anticipated, and keep the data in your own systems.
How quickly do logs arrive?
Each edge uploads a file every five minutes. After you switch delivery on, the first files appear within about ten minutes.
How long are log files kept?
Thirty days, then they are deleted automatically. Pull them into your own storage on a schedule to keep a longer history; cdnctl logs pull fetches only new files, so frequent runs are cheap.
Which time zone do the files use?
Files are grouped by UTC day. Each line carries its own timestamp in ISO 8601 with the offset, so your tools can convert it to any time zone.
Can I keep full visitor IPs out of the files?
Yes. With IP masking on, IPv4 addresses keep their /24 network and IPv6 addresses their /48, so the files contain no full visitor addresses.
Which domains does Logpush cover?
Every domain your CDN account serves through the edge. One switch on the account covers them all, cache hits and misses alike.