docs: Update on --limit-ledger-size and ledger disk usage#30002
docs: Update on --limit-ledger-size and ledger disk usage#30002steviez merged 1 commit intosolana-labs:masterfrom
Conversation
| The default value attempts to keep the blockstore (data within the rocksdb | ||
| directory) disk usage under 500GB. More or less disk usage may be requested | ||
| by adding an argument to `--limit-ledger-size` if desired. More information | ||
| about selecting a custom limit value is [available | ||
| here]https://github.com/solana-labs/solana/blob/aa72aa87790277619d12c27f1ebc864d23739557/core/src/ledger_cleanup_service.rs#L26-L37). | ||
|
|
||
| Note that the above target of 500 GB does not account for other items that | ||
| may reside in the `ledger` directory, depending on validator configuration. | ||
| These items may include (but are not limited to): | ||
| - Persistent accounts data | ||
| - Persistent accounts index | ||
| - Snapshots (both full and/or incremental) |
There was a problem hiding this comment.
The old comments don't call out that other stuff gets lumped in ledger. I called those out specifically as those are the largest items that I'm aware of that might end up under ledger.
That being said, a new operator would probably like to have some idea about the size of these. The aspects that make this tricky are that:
- The size of snapshots / accounts / accounts index will vary by cluster
- The size of above items all grow as number of accounts grow
- Other config options could make these vary (ie number of snapshots to retain)
As a result, I'm reluctant to put a static value as it will become out of date. Even if we choose a number for each cluster, those numbers will get out of date (even if we over-estimate).
Idk, maybe we should overestimate and have a recurring chore to update numbers on some interval. Curious to here other's thoughts
There was a problem hiding this comment.
I think it's good to clarify what is included in the 500 GB and what is not, but I think ultimately validator owners would like to know how much storage size is considered enough.
Do we also want to provide size estimations for the items that you list above?
There was a problem hiding this comment.
but I think ultimately validator owners would like to know how much storage size is considered enough.
Agreed, but again, I'm worried about this information becoming out of date. I think not providing the information is arguably better than having outdated information. I don't think there is anything we could point to that would allow someone to get current numbers for each cluster.
But, having something would be nice which is why I included two reviewers to brainstorm on what is otherwise a minor doc update 😉
The current documentation fails to call out that some items go under the ledger directory by default. If this is the case, then the stated 500 GB number will most certainly be exceeded.
8f0ca06 to
8681678
Compare
Pull request has been modified.
|
Per new merge rule, I need another ship it (first one dismissed). Any chance I can get a second look @brooksprumo & @yhchiang-sol ? |
The current documentation fails to call out that some items go under the ledger directory by default. If this is the case, then the stated 500 GB number will most certainly be exceeded. (cherry picked from commit a09b044)
…kport of #30002) (#30369) docs: Update on --limit-ledger-size and ledger disk usage (#30002) The current documentation fails to call out that some items go under the ledger directory by default. If this is the case, then the stated 500 GB number will most certainly be exceeded. (cherry picked from commit a09b044) Co-authored-by: steviez <steven@solana.com>
Problem
The current documentation fails to call out that some items go under the ledger directory by default. If this is the case, then the stated 500 GB number will most certainly be exceeded.
Summary of Changes
Add some additional details so that operators aren't immediately given false expectations about disk footprint of the ledger.
#26651 made it so space used by blockstore is kept in check better; but it is arguable that #23248 is still an issue given that with default args, ledger will grow past 500 GB with blockstore / accounts / accounts index / snapshots.
Fixes #