[Steem on RPi] Updates

etainclub (78)memberin Steemit Dev Group • 3 years ago (edited)

I have been trying to run a witness on a RPi 4 (8GB). Here are related posts regarding this issue.

--
I could not succeed on syncing block data even though I tried all the options that I can try.

But I will keep digging on this.

Here are some related links:

How to set up your Witness server with MIRA?

MIRA caches the data on RocksDB files rather than the random access memory, RAM. Which means you will require a higher storage with fast R/W speeds like SSDs for example. Instead of 64 GB of RAM.

I think RPi which has low RAM (8GB) should take the MIRA method.

Here is some info about MIRA:
https://github.com/steemit/steem/blob/master/doc/mira.md

When I tried replay, the log messages were

As @steemchiller suggested as following:

Maybe you can replay on a faster ARM machine with same OS and then just copy the data to the Pi.
I tried to sync on Apple M1 macbook air, but I could not run ./run.sh of steem-docker-ex. So I launched a docker container and executed steemd with --replay option. But it failed to sync too.

When I run replay on RPi4, the block_log.index file became very large which resulted in takes all the RPi's storage of 512GB. And the log messages were like this (when replaying):

------------------------------------------------------

            STARTING STEEM NETWORK

------------------------------------------------------
initminer public key: STM8GC13uCZbP44HzMLV6zPZGwVQ8Nt4Kji8PapsPiNq1BK153XTX
chain id: 0000000000000000000000000000000000000000000000000000000000000000
blockchain version: 0.23.0
------------------------------------------------------
2476125ms chain_plugin.cpp:468          plugin_startup       ] Starting chain with shared_file_size: 8589934592 bytes
2476125ms chain_plugin.cpp:571          plugin_startup       ] Replaying blockchain on user request.
2476126ms database.cpp:235              reindex              ] Reindexing Blockchain
2484048ms block_log.cpp:142             open                 ] Log is nonempty
2484049ms block_log.cpp:173             open                 ] Index is empty
2484049ms block_log.cpp:355             construct_index      ] Reconstructing Block Log Index...

I wonder what will happen if the storage is 1TB. This is the one that I did not try.

MIRA flavor of steemd fails to resync/replay while re-using old datadir
I found an issue having similar error message.
steem-0.23.1 repo
@ety001 introduced the repo.
https://github.com/hfang1989/steemhf/tree/hf0.23.1
The only difference between this 0.23.1 and 0.23.x of the official repo is not significant.
When starting only with blocklog file (no block_log.index, rockdb_xxxx folders), the log messages were the same as one before (cannot open database):

1901911ms database.cpp:192              open                 ] args.data_dir: /steem/witness_node_data_dir/blockchain args.shared_mem_dir: /steem/witness_node_data_dir/blockchain args.shared_file_size: 8589934592
1901912ms chain_plugin.cpp:608          plugin_startup       ] Error opening database. If the binary or configuration has changed, replay the blockchain explicitly using `--replay-blockchain`.
1901912ms chain_plugin.cpp:609          plugin_startup       ] If you know what you are doing you can skip this check and force open the database using `--force-open`.
1901912ms chain_plugin.cpp:610          plugin_startup       ] WARNING: THIS MAY CORRUPT YOUR DATABASE. FORCE OPEN AT YOUR OWN RISK.
1901912ms chain_plugin.cpp:611          plugin_startup       ]  Error: {"code":13,"name":"St13__ios_failure","message":"basic_ios::clear: iostream error","stack":[{"context":{"level":"warn","file":"raw.hpp","line":337,"method":"operator()","hostname":"","timestamp":"2021-08-10T14:31:41"},"format":"${what}: Error unpacking field ${field}","data":{"field":"props","what":"basic_ios::clear: iostream error"}},{"context":{"level":"warn","file":"raw.hpp","line":624,"method":"unpack","hostname":"","timestamp":"2021-08-10T14:31:41"},"format":"error unpacking ${type}","data":{"type":"steem::protocol::witness_set_properties_operation"}},{"context":{"level":"warn","file":"raw.hpp","line":337,"method":"operator()","hostname":"","timestamp":"2021-08-10T14:31:41"},"format":"Error unpacking field ${field}","data":{"field":"operations"}},{"context":{"level":"warn","file":"raw.hpp","line":624,"method":"unpack","hostname":"","timestamp":"2021-08-10T14:31:41"},"format":"error unpacking ${type}","data":{"type":"steem::protocol::signed_transaction"}},{"context":{"level":"warn","file":"raw.hpp","line":337,"method":"operator()","hostname":"","timestamp":"2021-08-10T14:31:41"},"format":"Error unpacking field ${field}","data":{"field":"transactions"}},{"context":{"level":"warn","file":"raw.hpp","line":624,"method":"unpack","hostname":"","timestamp":"2021-08-10T14:31:41"},"format":"error unpacking ${type}","data":{"type":"steem::protocol::signed_block"}},{"context":{"level":"warn","file":"block_log.cpp","line":264,"method":"read_block_helper","hostname":"","timestamp":"2021-08-10T14:31:41"},"format":"rethrow","data":{}},{"context":{"level":"warn","file":"block_log.cpp","line":336,"method":"read_head","hostname":"","timestamp":"2021-08-10T14:31:41"},"format":"rethrow","data":{}},{"context":{"level":"warn","file":"database.cpp","line":192,"method":"open","hostname":"","timestamp":"2021-08-10T14:31:41"},"format":"rethrow","data":{"args.data_dir":"/steem/witness_node_data_dir/blockchain","args.shared_mem_dir":"/steem/witness_node_data_dir/blockchain","args.shared_file_size":"8589934592"}}]}

When starting with --resync option, it removed block_log file and the following messages appeared:

$ steemd --data-dir=/steem/witness_node_data_dir --resync

186675ms witness_plugin.cpp:532        plugin_startup       ] Launching block production for 1 witnesses.
186675ms witness_plugin.cpp:543        plugin_startup       ] witness plugin:  plugin_startup() end
191641ms p2p_plugin.cpp:229            handle_block         ] Error when pushing block:
10 assert_exception: Assert Exception
item->num > std::max<int64_t>( 0, int64_t(_head->num) - (_max_size) ): attempting to push a block that is too old
    {"item->num":95,"head":1091,"max_size":22}
    fork_database.cpp:57 _push_block

    {}
    database.cpp:880 _push_block

    {"new_block":{"previous":"0000005ec36fa5413af6bf3dddd717223c119d75","timestamp":"2016-03-24T16:10:15","witness":"initminer","transaction_merkle_root":"0000000000000000000000000000000000000000","extensions":[],"witness_signature":"1f71b790d79b0ad1d2e63f073bb9ef5afc797352bb6c4ebbbbf2d0a4c7ddd7bd29512dfa6a5200a6bd32f317c61f358a523ee906588a0aa70f1ca108e05256cfef","transactions":[]}}
    database.cpp:764 operator()

I shared the following error messages with the blurt engineer (jacob).

He said it is a dns problem.

I guess there were messages of some like this:

 {"new_block":{"previous":"0181945ba5b0e549be8fb27ee5677c9e168a49c9","timestamp":"2021-08-17T11:52:24","witness":"bukowski","transaction_merkle_root":"0000000000000000000000000000000000000000","extensions":[],"witness_signature":"20d7984d721d69f6b6142b608aada9223b45f768b0a1bd576d2587c61a51b7c4b93a6965fe601fb3514c997914ff54b47156c594b8ac970fbf49e2cce1f2a76f1d","transactions":[]}}
    database.cpp:764 operator()
3143996ms p2p_plugin.cpp:231            handle_block         ] Rethrowing as graphene::net exception
3147129ms fork_database.cpp:43          push_block           ] Pushing block to fork database that failed to link: 0181945d155aed67fb69702a790bce76d2e05240, 25269341
3147130ms fork_database.cpp:44          push_block           ] Head: 1091, 000004433bd4602cf5f74dbb564183837df9cef8
3147130ms p2p_plugin.cpp:229            handle_block         ] Error when pushing block:
4080000 unlinkable_block_exception: unlinkable block
block does not link to known chain
    {}
    fork_database.cpp:64 _push_block

    {}
    database.cpp:880 _push_block

But the timestamp was 10-12 hours before of the current time. I am not sure when this happened. I guess this happened when replaying without removing rockdb_xxx folders. It seemed to catch recent blocks but the time difference persisted (10-12 hours difference all the time).
At that time, the block_log and the block_log.index files did not increase in size.

This is for now. Any advice and ideas are welcome.

See you later.

cc.
@steemcurator01

#rpi #steem #witness #steem-rpi #sync #dev

3 years ago in Steemit Dev Group by etainclub (78)member

$55.93

Sort:

Trending

[-]

steemchiller (74) 3 years ago

I never remove the block_log.index file when replaying and I've not had any issues yet. The --resync arg will start a full resync and the whole process will take much longer to complete, because it will also build the block_log from start by syncing each block.

When I run replay on RPi4, the block_log.index file became very large which resulted in takes all the RPi's storage of 512GB.

This is definitely not normal. The index file only contains the positions and lengths of the blocks in block_log. So it looks like steemd on ARM has a problem with reading the blocks from the log.