Run an archive liteserver
Run an archive liteserver node with MyTonCtrl
This guide describes how to set up an archive liteserver using MyTonCtrl, with ZFS for storage compression and snapshots.
An archive liteserver node stores the entire block history of the TON blockchain. For applications requiring access to historical data, such as blockchain explorers or indexers, running an archive liteserver node is the recommended approach.
Prerequisites
- A server meeting the minimal hardware requirements
- An OS meeting the software requirements
Step 1: Prepare environment
1.1 Minimal hardware requirements
Running an archive liteserver requires large storage and network capacity:
- 16-core CPU
- 128 GB RAM
- NVMe Gen4+ SSD storage (Enterprise grade preferred), sustaining at least 64,000 provisioned IOPS:
- at least 16 TB with ZFS
lz4compression enabled, or - at least 20 TB without compression
- at least 16 TB with ZFS
- 1 Gbit/s symmetric connectivity (both inbound and outbound), ~16 TB/month at peak load
- Fixed (static) public IP address
If non-Enterprise SSDs are used, Autonomous Power State Transition (APST) must be disabled on the SSD and the performance PCIe ASPM policy enabled at the system level:
echo performance | sudo tee /sys/module/pcie_aspm/parameters/policy1.2 OS and system requirements
- Ubuntu 22.04/24.04 LTS or Debian 11/12
- Python 3.10 or higher
- Open Files Limit must be set above 4,000,000
1.3 Subscribe to official channels
Subscribe and follow the announcements provided for liteservers in the following Telegram channels:
| Channel | Network |
|---|---|
@tonstatus | TON Mainnet |
@testnetstatus | TON Testnet |
1.4 Install ZFS and prepare volume
Archive nodes benefit from ZFS due to its native compression and snapshot capabilities.
1.4.1 Install ZFS
sudo apt update
sudo apt install -y zfsutils-linux1.4.2 Verify physical block size
For NVMe drives, ZFS pool sector size must align with the drive's physical block size to avoid performance degradation. Most modern NVMe drives use 4K blocks. Verify the physical block size:
# Replace nvme0n1 with the target device name
cat /sys/block/nvme0n1/queue/physical_block_sizeIf the result is 4096, the -o ashift=12 parameter must be used during pool creation.
1.4.3 Create a storage pool
Create a ZFS pool named data. Use -o ashift=12 for 4K blocks (standard for most NVMe drives):
# Replace <DISK> with the device identifier (e.g., /dev/nvme1n1)
sudo zpool create -o ashift=12 data <DISK>There,
<DISK>is the target disk device identifier, e.g.,/dev/nvme1n1;<DISK1>,<DISK2>, and<DISK3>are additional disk device identifiers.
Combining disks in a stripe (JBOD) increases the total capacity but also the risk of data loss: if a single drive fails, the entire pool is lost. If sufficient disks are available, consider using mirror or raidz for redundancy.
1.4.4 Enable compression
Enable lz4 compression to save disk space with minimal CPU overhead:
sudo zfs set compression=lz4 data1.4.5 Create dataset and mount point
Create the dataset for TON data and set the mount point to /var/ton-work:
sudo zfs create data/ton-work
sudo zfs set mountpoint=/var/ton-work data/ton-work1.5 Prepare the operator account
To create a dedicated operator user and switch to it before installing MyTonCtrl:
-
Create a non-root user:
# Create a non-root operator user sudo adduser <USERNAME> sudo usermod -aG sudo <USERNAME> -
Switch to the new operator account by reconnecting via SSH:
# Option 1: Reconnect using the standard port exit ssh <USERNAME>@<SERVER_IP>
1.6 Benchmark server performance
Before installing, verify that the server meets performance requirements. Inadequate disk or network performance is the most common cause of node instability.
1.6.1 Network latency
Check latency to TON beacon nodes. Expect approximately 50 milliseconds to the nearest beacon and up to 300 milliseconds to the farthest:
ping beacon-eu-01.toncenter.com -c 6
ping beacon-apac-01.toncenter.com -c 61.6.2 Disk IOPS
Install fio and run a random read/write benchmark:
sudo apt install -y fio
fio --randrepeat=1 --ioengine=psync --direct=1 --gtod_reduce=1 --name=tlstest --bs=4k --iodepth=1 --size=40G --readwrite=randrw --numjobs=1 --group_reporting --filename=/tmp/ton-testfile --time_based=1 --runtime=60 --refill_buffers --buffer_compress_percentage=0
rm /tmp/ton-testfileThe --refill_buffers and --buffer_compress_percentage=0 flags force incompressible data. Without them, LZ4 on the data/ton-work dataset collapses fio's zero-filled writes and reports IOPS far above what the archive import sustains.
The minimum acceptable result is 10,000 IOPS for both read and write operations. If disk performance falls below these thresholds, the liteserver may fail to keep up with network traffic. Upgrade storage before proceeding.
1.6.3 Network bandwidth
Verify network throughput with speedtest-cli:
sudo apt install -y speedtest-cli
speedtest-cliEnsure download and upload speeds meet the 1 Gbit/s requirement.
1.7 Harden server security
Apply security hardening steps before exposing the server to the network:
SSH hardening
Avoid locking yourself out
Disabling password login, changing the SSH port, and restricting access by Match Address can lock the operator out of a remote server. Keep the current SSH session open and confirm a new login succeeds in a second session before closing the first one.
Apply the following SSH configuration changes in /etc/ssh/sshd_config:
-
Enable key-based authentication and disable password login:
PasswordAuthentication no PubkeyAuthentication yes -
Disable root login:
PermitRootLogin no -
Change the default SSH port, e.g., to
2222:Port <SSH_PORT> -
Restrict SSH access to specific permitted IP addresses using the
Match Addressdirective:Match Address <ALLOWED_IP> AllowUsers <USERNAME>There,
<USERNAME>is the name of the operator user.
Restart the SSH service after changes:
sudo systemctl restart sshdFirewall configuration
Enable the firewall and allow only the SSH port. The node UDP port and liteserver port are added after installation in open the node UDP port and the liteserver port.
sudo apt install -y ufw
sudo ufw allow <SSH_PORT>
sudo ufw enable
sudo ufw statusAdditional security measures
-
Use a unique, strong password for the root user.
-
Set a GRUB bootloader password to prevent unauthorized boot modifications.
-
Enable Fail2ban for SSH brute-force protection:
sudo apt install -y fail2ban sudo systemctl enable fail2ban sudo systemctl start fail2ban -
Configure two-factor authentication for SSH using
libpam-google-authenticatoror a similar PAM module.
Step 2: Archive liteserver installation
The installation process consists of three stages (in total, this can take up to a week):
- Download historical blocks from TON Storage and install the archive liteserver
- Import downloaded data into the archive liteserver database
- Final synchronization of the archive liteserver
2.1 Download historical blocks from TON Storage and install the archive liteserver
This process can take from one to several days depending on the internet connection speed.
2.1.1 Install prerequisites and download MyTonCtrl installer
sudo apt update
sudo apt install -y curl wget git ca-certificates python3-pip
wget https://raw.githubusercontent.com/ton-blockchain/mytonctrl/master/scripts/install.sh2.1.2 Run archive liteserver installation
Run the installer from the operator account with sudo so it can create system users and services:
sudo -v && nohup sudo bash install.sh -m liteserver -n mainnet --archive > mytonctrl_installation.log 2>&1 &Installation runs in the background.
Monitor the progress using the following command:
tail -f mytonctrl_installation.logDuring the download process, the log contains entries like the following:
[info] 24.04.2026, 14:26:47.609 (UTC) <ThreadPoolExecutor-0_3>STARTING DOWNLOADING e88e7e805dfa16dc6e7d3864fcc00159a1ee70edde7b421428783c2164453875
[info] 24.04.2026, 14:27:07.611 (UTC) <ThreadPoolExecutor-0_3>DOWNLOADING e88e7e805dfa16dc6e7d3864fcc00159a1ee70edde7b421428783c2164453875 0% (0.0 / 4296.493726 MB), speed: 0.0 MB/s
[info] 24.04.2026, 14:27:07.623 (UTC) <ThreadPoolExecutor-0_2>DOWNLOADING 7683b6bc2c19e007fc6d363dc233ff9a405c0fee7533c2904c06943f759c155f 3% (130.766577 / 4295.81461 MB), speed: 15.078028 MB/s
[info] 28.04.2026, 11:43:00.687 (UTC) <ThreadPoolExecutor-0_1>DOWNLOADING fd41a820efc4f459bb1518c45fcc23023b0a95d9446c683705802bfb9d50a0c9 95% (4072.892954 / 4296.3534 MB), speed: 28.962779 MB/s
[info] 28.04.2026, 11:43:20.700 (UTC) <ThreadPoolExecutor-0_1>DOWNLOADED fd41a820efc4f459bb1518c45fcc23023b0a95d9446c683705802bfb9d50a0c9Upon successful completion of the installation, the following line appears in the log:
[5/5] Mytonctrl installation completed2.2 Import downloaded data into the archive liteserver database
This process starts automatically after installation and can take from one to several days depending on server performance.
Monitor the progress from the MyTonCtrl console. Open the console:
mytonctrlAt the MyTonCtrl> prompt, run:
MyTonCtrl> statusCheck the Local validator initial sync status field. The value indicates how old the last imported block was and should decrease over time.
2.2.1 Open the node UDP port and the liteserver port
At this stage, the node UDP port and liteserver port should be opened to make the archive liteserver available for syncing blocks from other nodes.
Identify the node UDP port and liteserver port from the config.json file:
sudo grep -A5 '"addrs"' -n /var/ton-work/db/config.json | grep '"port"' | head -1
sudo grep -A5 '"liteservers"' -n /var/ton-work/db/config.json | grep '"port"' | head -1Update security groups or configure ufw on bare-metal hosts:
sudo ufw allow <NODE_UDP_PORT>
sudo ufw allow <LITESERVER_PORT>
sudo ufw statusThere,
<NODE_UDP_PORT>is the UDP port of the validator engine;<LITESERVER_PORT>is the TCP port of the liteserver.
2.3 Final synchronization of archive liteserver
This process starts automatically after the importing process finishes and can take from one to several days depending on server performance.
Monitor the progress from the MyTonCtrl console. Open the console:
mytonctrlAt the MyTonCtrl> prompt, run:
MyTonCtrl> statusWhile initial sync continues, the Local validator initial sync status field reports how old the last imported block was, decreasing over time. Once initial sync completes, that line disappears and freshness is reported by the Local validator out of sync field. On a fully synchronized node, out-of-sync time stays below 20 seconds.
Step 3: Maintenance
3.1 Set up alerting
Set up alerting in MyTonCtrl to get a notification of critical issues with the archive liteserver. For more information, see MyTonCtrl private alerting bot.
3.2 Set up monitoring
Set up monitoring dashboards for RAM, disk, network, CPU usage, and other metrics.
It is critical to use the monitoring system to:
- monitor server stability;
- monitor synchronization parameters;
- check for memory leaks.
For system-level metrics, integrate Prometheus with node_exporter with MyTonCtrl.
For technical assistance, contact @mytonctrl_help_bot.
3.3 Perform software updates
Follow the @tonstatus channel, turn on notifications, and be prepared for urgent updates.
Before performing updates, create a ZFS snapshot of the data. This allows a quick rollback if the update process fails or corrupts the database.
sudo zfs snapshot data/ton-work@before-update-$(date +%Y-%m-%d)Update the node software and MyTonCtrl from the console. Open the console:
mytonctrlAt the MyTonCtrl> prompt, update MyTonCtrl to the tip of the master branch:
MyTonCtrl> update masterThe console exits when update finishes. Reopen it with mytonctrl and upgrade the TON node binaries to the tip of the master branch:
MyTonCtrl> upgrade masterThese commands check for new versions of MyTonCtrl and the TON node binaries, download them, and apply the updates. The update process may cause temporary node downtime as the binaries are replaced and services are restarted.
Once the update is verified as successful and the node is running correctly, delete the snapshot to reclaim disk space.
3.4 ZFS snapshots
ZFS creates snapshots for easy rollbacks if data corruption occurs.
Create a snapshot
Snapshots can be created under a unique identifier without stopping the node :
sudo zfs snapshot data/ton-work@<SNAPSHOT_NAME>List snapshots
To see all existing snapshots for the data/ton-work dataset:
sudo zfs list -t snapshot data/ton-workRoll back to a snapshot
Data loss risk
Rolling back to a snapshot overwrites all database changes made since the snapshot was created. Stop the validator service before performing the rollback:
# Stop the service
sudo systemctl stop validator.service
# Roll back to the snapshot
sudo zfs rollback data/ton-work@<SNAPSHOT_NAME>
# Start the service
sudo systemctl start validator.serviceDelete a snapshot
Delete the snapshot when it is no longer needed:
sudo zfs destroy data/ton-work@<SNAPSHOT_NAME>3.5 Archive ZFS snapshots
Creating a snapshot is instantaneous and occurs on the same physical disks where the data is stored. To protect against hardware failure, export the snapshots to external storage or a remote server using zfs send.
Export a snapshot to a file
Before exporting, estimate the size of the snapshot:
sudo zfs send -pc -nv data/ton-work@<SNAPSHOT_NAME>Expected output:
full send of data/ton-work@<SNAPSHOT_NAME> estimated size is 4.07T
total estimated size is 4.07TTo save the snapshot to a file, use the -c flag to preserve LZ4 compression. Without it, the file expands to the uncompressed size of the dataset:
# Ensure the destination directory exists and has enough free space!
sudo zfs send -c data/ton-work@<SNAPSHOT_NAME> > <BACKUP_PATH>Example destination path on the external storage: /mnt/backup/backup_ton_work.zfs
Exporting large datasets can take several hours. For example, a 4.07 TB snapshot may take approximately 2 hours to export, depending on disk throughput.
Transfer a snapshot via SSH
Stream a snapshot directly to a remote ZFS-enabled server:
sudo zfs send -c data/ton-work@<SNAPSHOT_NAME> | ssh <REMOTE_USER>@<REMOTE_HOST> "sudo zfs recv <REMOTE_POOL>/ton-work"There,
<REMOTE_USER>is the username on the remote host;<REMOTE_HOST>is an IP address or hostname of the remote host;<REMOTE_POOL>is the name of the remote ZFS pool.
Troubleshooting
Monitor import logs
To see detailed logs of the block import process, increase the log verbosity from the MyTonCtrl console. Open the console:
mytonctrlAt the MyTonCtrl> prompt, run:
MyTonCtrl> installer set_node_argument --verbosity 3Then follow the log file from a separate terminal:
tail -f /var/ton-work/log*Expected log entries:
[ 2][t49][2025-01-01 00:00:00.632][import-db-slice-local.cpp:629][!archiveimport] Imported archive in 2.75s : mc_seqno=761229 shard_seqno=761229Set verbosity back to 1 after checking logs to avoid excessive disk I/O overhead. At the MyTonCtrl> prompt, run:
MyTonCtrl> installer set_node_argument --verbosity 1Performance issues
Logs containing "Importing archive for masterchain seqno #... from net" accompanied by timeout errors indicate insufficient storage performance. Ensure the disk meets the IOPS requirements listed in Minimal hardware requirements.
To verify disk and system performance, run the built-in mytonctrl benchmark:
-
Stop the validator service, since the benchmark refuses to run while it is active:
sudo systemctl stop validator.service -
Open the MyTonCtrl console:
mytonctrlAt the
MyTonCtrl>prompt, run:MyTonCtrl> benchmarkThe benchmark spins up a local test network and requires
uv. Ifuvis not installed, the console prompts to install it. For stable liteserver operation, the reportedAvg TPSandAvg blocks/sshould each reach at least 70% of their expected values. -
Restart the validator service once the benchmark finishes:
sudo systemctl start validator.service
Support
For technical assistance, join the official support channel: @ton_node_help.