To many secret calls cause a memeory explosion

[fuerststat]

I'm doing ETL tasks using ssecret for DNS hiding in Statements.
The Memory (RAM) which is allocated by this "INSERT" statements cannot be freed by GC until I kill my R session (RAM sums up to more than 24-gig during the proces, not using secrets max is 2gig).
Can anyone explain this behaviour to me?

KR, Leo

[gaborcsardi]

Maybe a memory leak. Which backend are you using?

[fuerststat]

R3.4.3 on R Studio Server running on Ubuntu 16.
I thought, maybe the decrypted secret_raw_string (means object or variable) is marked permanently as "active". So the GC would never be able to collect it.
But it's just guessing!?

KR, Leo

[gaborcsardi]

Are you using libsecret or environment variables?

[fuerststat]

i prefer neither and call the valut for each statement.
Storing the value in a env-variable, prevents the mem-explosion.
nevertheless i would prefere the vault solution, as also people "debugging" the code are not able to access the secrets

[gaborcsardi]

Seems like a memory leak in the openssl package: jeroen/openssl#79

[gaborcsardi]

Seems like it is not read_key after all, so it must be something else.