Problems reconstructing an archive node

Question

I've encountered several problems in trying to reconstruct an archive node. I've tried several machines and continue to encounter errors. I've tried both reconstructing an archive node by importing a full snapshot (2nd machine) and tried reconstructing an archive node from an existing full node (1st machine). The 1st machine is 1tb ssd drive with 8gb ram, modern Intel cpu and 2nd machine is a 1tb nvme drive, 32 gb ram, modern AMD cpu. The reconstruct process works great for both machines initially. The error occurs when the ".tezos-node" directory reaches > 100 Gb.

First Machine

1. The command "nohup tezos-node reconstruct &" gave the following error:

   cat nohup.out Mar 7 17:03:47.021 - snapshot: Importing data from snapshot file ./BMbyDQVNNmWRncRL87xwFCgY5iDBVSuhrz5MiYWFjFVgXfUt7Br.full Mar 7 17:03:47.021 - snapshot: Retrieving and validating data. This can take a while, please bear with us Mar 7 17:42:07.171 - snapshot: Setting current head to block BMbyDQVNNmWRncRL87xwFCgY5iDBVSuhrz5MiYWFjFVgXfUt7Br Mar 7 17:42:08.384 - snapshot: Setting history-mode to full Mar 7 17:42:08.888 - snapshot: Successful import from file ./BMbyDQVNNmWRncRL87xwFCgY5iDBVSuhrz5MiYWFjFVgXfUt7Br.full Mar 24 22:33:43.725 - snapshot: Starting reconstruct from genesis Mar 24 22:33:43.726 - snapshot: Starting reconstruct toward the predecessor of the current head (BLbiPyETrSNa69bWHvxvmzChzvNyZNycbztEZxvn8xXsCrB1q3T) Mar 24 22:33:43.729 - snapshot: Enumerating all blocks to reconstruct Mar 24 22:46:07.703 - snapshot: Setting history-mode to archive (/home/xxxx/.opam/for_tezos/bin/tezos-node) Exit: exit because of uncaught exception: (Failure "MDB_MAP_FULL: Environment mapsize limit reached")

Second Machine

1. The command "nohup tezos-node snapshot import snapshot.full --reconstruct &" gave the following error:

   cat nohup.out Mar 19 17:35:57.964 - snapshot: Importing data from snapshot file snapshot.full Mar 19 17:35:57.964 - snapshot: You may consider using the --block <block_hash> argument to verify that the b(/usr/bin/tezos-node) Exit: exit because of uncaught exception: (Failure "MDB_BAD_TXN: Transaction must abort, has a child, or is invalid")lock imported is the one you expected Mar 19 17:35:57.964 - snapshot: Retrieving and validating data. This can take a while, please bear with us Mar 19 17:48:20.486 - snapshot: Setting current head to block BLEz5LuE69e5LYL32L6MwvQeYrdqJbXM8itMsKJuQbNV6q5SJ4g Mar 19 17:48:21.333 - snapshot: Setting history-mode to archive (/usr/bin/tezos-node) Exit: exit because of uncaught exception: (Failure "MDB_BAD_TXN: Transaction must abort, has a child, or is invalid") Mar 24 20:04:57.867 - snapshot: Importing data from snapshot file snapshot.full Mar 24 20:04:57.867 - snapshot: You may consider using the --block <block_hash> argument to verify that the block imported is the one you expected Mar 24 20:04:57.867 - snapshot: Retrieving and validating data. This can take a while, please bear with us Mar 24 20:19:27.140 - snapshot: Setting current head to block BLEz5LuE69e5LYL32L6MwvQeYrdqJbXM8itMsKJuQbNV6q5SJ4g Mar 24 20:19:28.060 - snapshot: Setting history-mode to archive (/usr/bin/tezos-node) Exit: exit because of uncaught exception: (Failure "MDB_BAD_TXN: Transaction must abort, has a child, or is invalid")

Answer 1

This error is related to the LMDB library which is used for storing the block's history (aka "the store").

Regarding the first machine, it is a bit strange as the mapsize value was set to avoid reaching this "high" limit. Furthermore, as the mainnet chain as reached a higher level than the snapshot that you are importing (snapshot: 1,273,857 vs current head: 1,408,297), this error should have appeared before for other people.

Regarding the second machine, I seem to be some data corruptions which is also quite strange.

Can you give additional information regarding the Operating System on which your machines are running?